CN111639843A - Provider selection method and device based on residual error neural network - Google Patents

Abstract

The invention discloses a method and device for selecting a supplier based on a residual neural network. The method for selecting a supplier based on a residual neural network includes: acquiring the collected preset evaluation parameters of the supplier; inputting the evaluation parameters into Obtain the evaluation results of the supplier in the trained supplier evaluation model, so as to select suppliers according to the evaluation results of multiple suppliers, wherein the supplier evaluation model uses the evaluation parameters marked with the evaluation results as training The data is obtained by training a residual neural network. The present invention can accurately evaluate each supplier, and then select the supplier according to the evaluation result of each supplier, thereby realizing the technical effect of accurate and rapid selection of suppliers by the enterprise.

Description

Method and device for supplier selection based on residual neural network

technical field

The present invention relates to the technical field of deep learning, and in particular, to a method and device for selecting a supplier based on a residual neural network.

Background technique

With the increasing marketization and diversification of procurement methods, the selection of suppliers who can provide high-quality products plays an important role in ensuring the stable development of enterprises. At present, when evaluating suppliers, the evaluation index is too single and the evaluation method is too subjective, which makes it difficult to make a comprehensive and objective evaluation of suppliers, and it is difficult for enterprises to select suppliers more objectively.

SUMMARY OF THE INVENTION

In order to solve at least one technical problem in the above background technology, the present invention proposes a method and device for selecting a supplier based on a residual neural network.

In order to achieve the above object, according to an aspect of the present invention, there is provided a residual neural network-based supplier selection method, the method comprising:

Obtain the preset evaluation parameters of the collected suppliers;

Input the evaluation parameters into the trained supplier evaluation model to obtain the evaluation results of the suppliers, so as to select suppliers according to the evaluation results of multiple suppliers, wherein the supplier evaluation model is marked with a good evaluation The evaluation parameters of the results are used as training data and obtained by using residual neural network training.

Optionally, the residual neural network-based supplier selection method further includes:

Acquiring training data, wherein the training data includes: evaluation parameters used for model training and evaluation results corresponding to the marked evaluation parameters used for model training;

The supplier evaluation model is obtained by continuously performing model training according to the training data and the preset residual neural network, wherein, during model training, the evaluation parameters used for model training are input into the residual In the neural network, the parameters of the residual neural network are adjusted so that the output of the residual neural network approximates the evaluation result corresponding to the evaluation parameters used for model training.

Optionally, inputting the evaluation parameters into the trained supplier evaluation model to obtain an evaluation result of the supplier, specifically including:

The first layer network of the supplier evaluation model weights and biases the evaluation parameters through the first layer network weight and the first layer network bias, and the result is input to the first activation function in the first layer network , the output of the first layer network is obtained;

The second-layer network of the supplier evaluation model weights and biases the output of the first-layer network through the second-layer network weight and the second-layer network bias, so as to obtain the output of the second-layer network;

The output of the second-layer network is superimposed with the evaluation parameters, and the superimposed result is input into the second activation function to obtain the evaluation result of the supplier.

Optionally, the evaluation parameters include: at least one of quality level parameters, technical level parameters, supply capacity parameters, procurement cost parameters, financial risk level parameters, and service level parameters.

Optionally, the quality level parameter includes: at least one of product qualification rate, excellent batch rate and quality assurance capability parameter; the technical level parameter includes: product research and development capability parameter, product design capability parameter and independent production capability parameter At least one of; the supply capacity parameter includes: at least one of on-time delivery rate, logistics efficiency, supply flexibility and procurement cost; the procurement cost parameter includes: price level parameter, price rationality parameter, after-sales guarantee cost, at least one of maintenance cost and price reduction capability parameters; the financial risk level parameter includes: whether there is a financial risk index; the service level parameter includes: response to customer demand rate.

In order to achieve the above object, according to another aspect of the present invention, there is provided a residual neural network-based supplier selection device, the device comprising:

an evaluation parameter obtaining unit, used for obtaining the collected preset evaluation parameters of the supplier;

A supplier selection unit, configured to input the evaluation parameters into the trained supplier evaluation model to obtain the evaluation result of the supplier, and then select suppliers according to the evaluation results of multiple suppliers, wherein the supplier The quotient evaluation model is obtained by using the evaluation parameters marked with the evaluation results as training data and using residual neural network training.

Optionally, the apparatus for selecting suppliers based on residual neural network further includes:

a training data acquisition unit, configured to acquire training data, wherein the training data includes: evaluation parameters used for model training and evaluation results corresponding to the marked evaluation parameters used for model training;

A model training unit, configured to continuously perform model training according to the training data and the preset residual neural network to obtain the supplier evaluation model, wherein, during model training, the evaluation parameters used for model training are used for model training. Input into the residual neural network, and adjust the parameters of the residual neural network so that the output of the residual neural network approximates the evaluation result corresponding to the evaluation parameter used for model training.

Optionally, the supplier selection unit specifically includes: a first-layer network, a second-layer network, and a processing module;

The first-layer network is used for weighting and biasing the evaluation parameters through the first-layer network weight and the first-layer network bias, and inputting the result into the first activation function to obtain the first-layer the output of the network;

The second-layer network is used for weighting and biasing the output of the first-layer network through the second-layer network weight and the second-layer network bias, so as to obtain the output of the second-layer network;

The processing module is configured to superimpose the output of the second layer network and the evaluation parameter, and input the superimposed result into the second activation function to obtain the evaluation result of the supplier.

In order to achieve the above object, according to another aspect of the present invention, a computer device is also provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, the processor executing the computer The program implements the steps in the above-mentioned residual neural network-based supplier selection method.

In order to achieve the above object, according to another aspect of the present invention, a computer-readable storage medium is also provided, where the computer-readable storage medium stores a computer program, and when the computer program is executed in a computer processor, the above-mentioned based on Steps in a vendor selection method for residual neural networks.

The beneficial effects of the present invention are as follows: the embodiment of the present invention can accurately evaluate each supplier by setting evaluation parameters capable of evaluating various capabilities of suppliers, and training a supplier evaluation model based on the residual neural network. The evaluation results of each supplier are used for supplier screening/selection, which realizes the technical effect of accurate and rapid supplier selection.

Description of drawings

In order to illustrate the embodiments of the present invention or the technical solutions in the prior art more clearly, the following briefly introduces the accompanying drawings that need to be used in the description of the embodiments or the prior art. Obviously, the drawings in the following description are For some embodiments of the present invention, for those of ordinary skill in the art, other drawings can also be obtained according to these drawings without creative efforts. In the attached image:

1 is a flowchart of a method for selecting a supplier based on a residual neural network according to an embodiment of the present invention;

Fig. 2 is the training flow chart of the supplier evaluation model of the embodiment of the present invention;

Fig. 3 is the flow chart that the supplier evaluation model of the embodiment of the present invention calculates the evaluation result;

4 is a schematic diagram of a residual neural network according to an embodiment of the present invention;

5 is a first structural block diagram of a residual neural network-based supplier selection device according to an embodiment of the present invention;

FIG. 6 is a second structural block diagram of the apparatus for selecting a supplier based on a residual neural network according to an embodiment of the present invention;

FIG. 7 is a schematic diagram of a computer device according to an embodiment of the present invention.

Detailed ways

In order to make those skilled in the art better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only Embodiments are part of the present invention, but not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

As will be appreciated by one skilled in the art, embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) having computer-usable program code embodied therein.

It should be noted that the terms "comprising" and "having" in the description and claims of the present invention and the above-mentioned drawings, as well as any variations thereof, are intended to cover non-exclusive inclusion, for example, including a series of steps or units The processes, methods, systems, products or devices are not necessarily limited to those steps or units expressly listed, but may include other steps or units not expressly listed or inherent to such processes, methods, products or devices.

It should be noted that the embodiments of the present invention and the features of the embodiments may be combined with each other under the condition of no conflict. The present invention will be described in detail below with reference to the accompanying drawings and in conjunction with the embodiments.

1 is a flowchart of a method for selecting a supplier based on a residual neural network according to an embodiment of the present invention. As shown in FIG. 1 , the method for selecting a supplier based on a residual neural network in this embodiment includes steps S101 to S102 .

In step S101, the collected preset evaluation parameters of the supplier are acquired.

Step S102, inputting the evaluation parameters into the trained supplier evaluation model to obtain the evaluation results of the suppliers, so as to select suppliers according to the evaluation results of multiple suppliers, wherein the supplier evaluation model is based on the evaluation results of multiple suppliers. The evaluation parameters marked with the evaluation results are used as training data and obtained by using residual neural network (ResNet) training.

In an optional embodiment of the present invention, when selecting suppliers, the present invention inputs the collected evaluation parameters of multiple suppliers into the trained supplier evaluation model respectively, and obtains the evaluation results of each supplier, and then according to The evaluation results of each supplier are used for supplier screening or selection. In an optional embodiment of the present invention, the evaluation result of the supplier may be expressed in the form of a score, preferably a percentage system. In other optional embodiments of the present invention, the evaluation results of the suppliers may also be expressed in the form of ratings, such as "excellent", "good", "medium", "poor" and so on. In an optional embodiment of the present invention, when suppliers are selected according to the evaluation results of each supplier, the supplier with the best evaluation result may be selected, or the supplier with the top N (integer greater than 1) ranking in the evaluation result may be selected as the supplier to be selected.

In an optional embodiment of the present invention, the present invention first establishes the evaluation parameters to be considered in supplier selection, and determining the evaluation parameters needs to follow the principles of target clarity, adaptability, scientificity, and operability.

The principle of clarity of goals: The selected supplier evaluation parameters should be clear, and should be able to truly reflect the supplier's evaluation content, so that the scope of evaluation covers a large area, and flexibly distinguish between relevant evaluation parameters and irrelevant evaluation parameters, so as to avoid irrelevant evaluation parameters. The evaluation parameters affect the implementation of relevant decisions.

Adaptability principle: Different industry types have different strategies and business characteristics at different development stages, so supplier evaluation parameters are also different. The goal and business philosophy pursued by its own development, and the parameters that are suitable for itself are selected.

Scientific principle: The selected evaluation parameters should be able to truly reflect the actual situation of the supplier. If the evaluation parameter is too large, it cannot reflect the whole of the supplier; if it is too small, the actual level of the supplier cannot be clearly analyzed. Therefore, the selected evaluation parameters must follow the scientific principle.

The principle of operability: the selection of evaluation parameters for suppliers should be feasible, that is to say, when selecting and evaluating suppliers according to the set evaluation parameters, the selected parameters should be refined, and the data of each parameter should be easily obtained. , for the quantitative parameters in the parameters, it should be possible to obtain data through existing materials, sample surveys, etc., and for the qualitative parameters in the parameters, it should be able to be quantified by methods such as expert scoring. The calculation method of each parameter should be simple and feasible, and the calculation process should be standardized and normalized, thereby reducing the difficulty of evaluating suppliers.

In an optional embodiment of the present invention, the present invention determines six first-level evaluation parameters according to the above principles, which are: quality level parameters, technical level parameters, supply capacity parameters, procurement cost parameters, financial risk level parameters, and service level parameters , and each first-level evaluation parameter also includes several second-level evaluation parameters. These evaluation parameters are described below.

1. Quality level parameters

The requirement for supply quality has always been one of the most important parameters in the supplier selection evaluation parameter system. In an optional embodiment of the present invention, the quality level parameters include: product qualification rate, excellent batch rate, and quality assurance capability parameters. The present invention evaluates the quality level of suppliers from three aspects: product qualified rate, excellent batch rate and quality assurance capability.

Rate of qualified products

The qualified rate of products refers to the proportion of qualified products such as raw materials or equipment supplied by suppliers to the total number of purchases in a certain period. The total amount is Q. Among the Q products, the number of qualified products is q, then the pass rate p of this batch of products is:

If multiple batches of products are purchased during the evaluation period, the weighted average method is used to calculate the pass rate, that is, the overall pass rate P of the product is:

Among them, p ₁ , p ₂ , ..., p _n are the qualification rates of the 1st, 2nd, ..., n batches of products, respectively.

Good batch rate

In a certain period of time, the percentage of good product batches purchased from suppliers to the total purchased batches is called the good batch rate, where the good batch refers to the batches with a product qualification rate greater than 95% in a single purchase Second-rate. Assuming that in one evaluation cycle, the total number of ordered product batches is N, and the number of good batches is n, then the good batch rate G is:

Quality Assurance Capability Parameters

Quality assurance capability refers to the quality management means and methods possessed by suppliers. These means and methods include Total Quality Management, quality process control, ISO series certification and so on.

The evaluation of the supplier's quality assurance capability is a complex process, and many factors need to be considered. The present invention provides a relatively simple evaluation method for the quality assurance capability. The present invention divides the supplier's quality assurance capability into excellent, good, and There are four grades of qualified and unqualified, which are represented by a score of 0-100 respectively. Among them, the suppliers with excellent quality assurance ability have passed ISO certification, and the supplier enterprise has a complete overall quality management system and quality management standards. The division of quality responsibilities is clear, the rewards and punishments are appropriate, the quality management work has achieved remarkable results, and the score is set to be greater than or equal to 85 points; suppliers with good quality assurance capabilities have a comprehensive quality management system, and suppliers have internal quality management standards. The quality management training is sufficient, and the score is between 70 and 85 points; the quality assurance ability of qualified suppliers is general, the quality management standards are not standardized, the quality awareness training for employees is insufficient, and the quality The responsibility mechanism is missing, and the score is between 60 and 75; the quality assurance ability is unqualified suppliers lack quality management, no relevant quality qualification certification, lack of awareness of quality management, no quality responsibility mechanism, and the score is less than 60 point.

2. Technical level parameters

The technical level requirement has always been one of the key considerations when companies in various industries choose to evaluate suppliers. For purchasing enterprises, in order to comply with the trend of informatization and scientificization, the technical level of suppliers is also becoming higher and higher, and the ability to manufacture high-tech products has increasingly become a key factor for suppliers to obtain core competitiveness. . In an optional embodiment of the present invention, the technical level parameters include: product research and development capability parameters, product design capability parameters, and independent production capability parameters.

Product R&D Capability Parameters

Generally speaking, the product R&D capability of an enterprise is mainly reflected in the level of R&D investment, R&D output, and cumulative R&D results. It is relatively abstract, and the measurement cost of this parameter is relatively high. Therefore, when an enterprise inspects the research and development level of a new product of a supplier, it can adopt reasonable and feasible parameters according to the needs of the enterprise itself. Assuming that the supplier's R&D investment in the assessment period is r million yuan, and the total profit during this period is R million yuan, then the supplier's product R&D capability parameter d is:

This parameter can measure the supplier's investment in product research and development, and is an important parameter to measure the technical degree of cooperation between suppliers and core enterprises in the supply chain environment.

Product Design Capability Parameters

Generally speaking, the more people engaged in product design, the more product design capabilities the supplier has to a certain extent. The supplier's product design capability can be measured by the proportion of product design personnel, which refers to the proportion of technical personnel engaged in product design to all professional and technical personnel.

Independent production capacity parameters

This parameter is used to measure the ability of the supplier to manufacture and supply the required raw materials and so on. Due to the uncertainty and urgency of specific product demand, suppliers are required to have certain independent production capacity when evaluating and selecting suppliers. The independent production capacity adopts a 0-100 scoring system, from no independent production capacity at all to complete production capacity to score suppliers.

Three, supply capacity parameters

The measurement of the supplier's supply capacity is one of the factors that many companies must consider when selecting suppliers. An excellent supplier must have a strong supply capacity, because in the market economy environment, the order demand faced by enterprises is often uncertain. Whether suppliers can supply raw materials in a timely and efficient manner is very important for enterprises to seize market opportunities and win customer satisfaction. Parameters such as purchasing cost are measured.

On-time delivery rate

The on-time delivery of suppliers is premised on the timely and accurate supply of raw materials and equipment, which is a major challenge to the on-time delivery capability of suppliers. A supplier's on-time delivery rate is measured by the number of batches delivered on time in a given period as a percentage of the total number of batches.

Logistics efficiency

The logistics efficiency of a supplier can be measured by the delivery lead time. Generally speaking, the longer the delivery lead time of the supplier, the stronger the response ability to the demand for raw materials and equipment, and the higher the logistics efficiency. The logistics efficiency of the supplier can be measured by the ratio of the number of deliveries in the lead period to the total number of times. Formula (2-5) can be used to calculate the logistics efficiency of the supplier, where s represents the number of deliveries in the lead period, and S represents the number of deliveries in the lead period. The total number of deliveries, u represents the logistics efficiency.

Supply flexibility

A supplier's supply flexibility measures its responsiveness to changes in orders. Use h to represent supply flexibility, C to represent the number of original orders, and c to represent the number of increased or decreased orders, then:

4. Purchase cost parameters

Procurement cost is a problem that enterprises must face to obtain higher profits. The cost factor of equipment suppliers will directly affect the company's ability to reduce costs, and even affect the company's development strategy. In an optional embodiment of the present invention, the procurement cost mainly includes a price level parameter, a price rationality parameter, an after-sale guarantee cost, a maintenance cost, a price reduction capability parameter, and other incentives. The purchase cost parameter mainly measures the comprehensive cost paid by the enterprise to purchase equipment and the ability to convert the cost into profit.

price level parameter

In the existing supplier selection evaluation parameter system, the supplier's price level is an important evaluation standard, and the price that the supplier can provide is an important factor that directly determines the cost of the enterprise. Therefore, the price level must be considered in the supplier selection evaluation. , which is crucial for corporate procurement. The price level parameter is a quantitative parameter, which is directly reflected by its purchase price.

Price Reasonability Parameters

Price reasonableness refers to the degree of difference between the quotation of a certain equipment supplier and the average price of the equipment in the market. For equipment quotations that are higher than the average price level and have large differences, it is necessary to conduct business negotiations with the equipment supplier after re-approval. Although the economic law of commodity value can analyze the price to reflect the value, under the influence of various factors such as market demand, the price will fluctuate to a certain extent around the value. The rationality of the purchase price can help enterprises to use market rules to purchase equipment products with high cost performance.

Purchasing companies can compare suppliers’ quotations with the price levels of similar products in the market. For some raw material prices that are difficult to obtain directly in the open market, purchasing companies can make adjustments based on historical purchase prices combined with current price levels and suppliers’ reasonable profit margins. Sure.

In the research of the present invention, the reasonableness of the supplier's price is measured by the ratio of the supplier's quotation to the average price of similar products in the market. It is assumed that P represents the average market price, p represents the supplier's quotation, and a represents the supplier's price rationality parameter. The specific calculation formula is:

After-sale guarantee cost

The cost of after-sales protection refers to the total cost spent by the enterprise when a series of non-supplier responsibility problems occur in the equipment during the after-sales service period of the equipment. This parameter can reflect the continuous potential cost output after the end of the procurement process, which is conducive to mining hidden costs and making them explicit. The after-sale guarantee cost is a quantitative parameter, which can be expressed by the average value of the after-sale guarantee costs reported by previous cooperation or other customers of the equipment.

Maintenance costs

Maintenance cost refers to the sum of expenses incurred by an enterprise in hiring equipment engineers or deploying relevant personnel when quality problems occur in equipment products. This parameter can not only reflect the frequency of equipment product failures, but also comprehensively evaluate the continuous cost output required for post-purchase failure maintenance.

_CR : maintenance costs, mainly including labor costs, material costs, machinery usage costs, other direct costs and indirect costs; _TB : downtime; _TP : production time.

Price reduction ability parameter

Price competitiveness refers to the competitiveness of suppliers' current quotations, while price reduction space refers to the competitiveness of suppliers' future quotations. Different suppliers have different room for price reduction in the future. This mainly depends on two points. One is the characteristics of the suppliers themselves. For large suppliers, with the deepening of cooperation, they will generate scale benefits, reduce the cost of unit products, and bring the possibility of price reduction. Suppliers are often unable to obtain marginal benefits of production; the second is the dependency relationship with suppliers. With the deepening of cooperation, some suppliers rely more and more on purchasing companies. Important customers may offer better prices, while businesses that are less reliant on sourcing companies may not offer larger price concessions.

The price reduction ability parameter can be quantified by the gap between the supplier's price reduction and the purchasing company's target price reduction. The smaller the gap, the better the equipment supplier's price reduction ability can achieve the expected goal. The specific calculation process is as follows:

Among them, D _A is the actual reduction rate, and D _G is the target reduction rate.

Other offers

Other preferential policies refer to preferential policies provided by equipment suppliers for enterprises beyond the scope of the above assessment, such as the actual performance of parts price reduction, more purchase feedback, payment concessions, etc. This parameter mainly evaluates the preferential policies of equipment suppliers that are not involved in the above cost parameters. Under the above strict parameter evaluation, suppliers with other preferential plans will also be preferred. This parameter can take into account various cost factors, so as to evaluate each supplier more fairly and impartially.

Other discounts mainly rely on the quantification of the total discounted amount of discounts, and discount and sum up various advantageous policies provided by suppliers as the basis for the evaluation of this parameter.

V. Financial Risk Level Parameters

In the selection and evaluation parameters of suppliers, the financial situation of the supplier is one of the parameters to be considered. This is because on the one hand, the supplier needs sufficient financial support for the production of raw materials or equipment. If the supplier does not have a strong financial Ability, it is difficult to imagine that it will complete the supply of raw materials and equipment with quality and quantity. On the other hand, in the measurement of general companies, financial ability is an indispensable parameter, and the financial status of a company is directly determined. Whether the cooperation with it can continue in a long-term and stable manner. Among the current evaluation parameters, there are many evaluation parameters for the supplier's financial capability, and the present invention selects whether there is a financial risk index, which is simple and clear. Set 0, 1 to represent whether the supplier is at risk. 0 means the supplier has no financial risk, 1 means the supplier has financial risk.

6. Service Level Parameters

The supplier's service level mainly refers to the supplier's ability to meet customer needs in a timely manner, including the degree of satisfaction to customers before the signing of the supply order, during the signing of the supply order, and the entire supply cycle after the order takes effect. The present invention measures the supplier's service level by responding to the customer demand rate. The higher the efficiency of the supplier in responding to the customer's demand, the higher the service level. The response to customer demand rate is obtained through customer scoring. The customer chooses a value between 0 and 100. This value also reflects the customer satisfaction in the entire supply process. For suppliers who have historical cooperation, this value can be obtained by the relevant procurement department of the enterprise based on the actual situation of the supplier; for suppliers who have not yet cooperated, it is necessary to fully consider the pre-sale, in-sale and after-sale service terms provided by them. based on the evaluation.

The present invention builds a relatively complete evaluation parameter system through literature research and combined with the actual situation, including 6 first-level evaluation parameters and 17 second-level evaluation parameters. The system covers almost all evaluation dimensions, so that the method of the present invention can evaluate suppliers more accurately and comprehensively, thereby improving the accuracy and comprehensiveness of supplier selection.

FIG. 2 is a training flow chart of a supplier evaluation model according to an embodiment of the present invention. As shown in FIG. 2 , in an optional embodiment of the present invention, the specific training process of the supplier evaluation model in step S102 above includes steps S201 and S202.

Step S201, acquiring training data, wherein the training data includes: evaluation parameters used for model training and evaluation results corresponding to the marked evaluation parameters used for model training.

In the embodiment of the present invention, a certain amount of training data is generated, and the training data is divided into a training set and a verification set, the training set is used for model training, and the verification set is used to verify the effect of the model.

Step S202, continuously perform model training according to the training data and the preset residual neural network to obtain the supplier evaluation model, wherein, when performing model training, the evaluation parameters used for model training are input into the supplier evaluation model. In the residual neural network, the parameters of the residual neural network are adjusted so that the output of the residual neural network approximates the evaluation result corresponding to the evaluation parameters used for model training.

FIG. 4 is a schematic diagram of a residual neural network according to an embodiment of the present invention. As shown in FIG. 4 , in an optional embodiment of the present invention, the residual neural network of the present invention includes a first-layer network and a second-layer network. The processing of the input data x by the difference neural network includes the following steps:

The data x is input into the first-layer network, and after weighting, biasing and activation function operations, the output F(x), F(x) is the output of the first-layer network, as shown in the following formula:

F(x)=R(xω ₁ +b ₁ )

Among them, ω ₁ is the weight of the first layer network; b ₁ is the bias of the first layer network; R() is the ReLU activation function.

Input the output F(x) of the first-layer network into the second-layer network, and output F ₁ (x) after weighting and biasing operations. F ₁ (x) is the output of the second-layer network, as shown in the following formula :

F ₁ (x)=F(x)ω ₂ +b ₂

Among them, ω ₂ is the weight of the second-layer network; b ₂ is the bias of the second-layer network.

Input F ₁ (x)+x into the activation function ReLU, output F ₂ (x), F ₂ (x) is the output of the residual neural network, as shown in the following formula:

F ₂ (x)=R(F ₁ (x)+x)

In an optional embodiment of the present invention, F ₁ (x) and x must have the same dimension to perform element-by-element addition operation. When F ₁ (x) and x have different dimensions, x can be transformed into a parameter with the same dimension as F ₁ (x) by multiplying x by the weight ω, as described in the following formula:

F ₂ (x)=R(F ₁ (x)+xω)

Combined with the processing steps of the above-mentioned residual neural network for the input data x, when performing model training, the evaluation parameters used for model training are taken as the data x in the above steps, and the marked evaluation parameters used for model training correspond to The evaluation result is denoted as H(x). During model training, the evaluation parameter x used for model training is input into the residual neural network, and the parameters of the residual neural network are continuously adjusted so that the output of the residual neural network is close to the marked one for the model. The evaluation result H(x) corresponding to the training evaluation parameters. In an optional embodiment of the present invention, the parameters of the residual neural network include: the weight of the first-layer network, the bias of the first-layer network, the weight of the second-layer network, the bias of the second-layer network, and the activation function of each parameter. In the embodiment of the present invention, the present invention conducts model training through a large amount of training data, and verifies the effect of the trained model in combination with the verification data, and finally obtains the optimal parameters of the residual neural network, which has the optimal parameters at this time. The residual neural network is the supplier evaluation model of the present invention.

It can be seen from the above model training process that the residual neural network of the present invention can convert the identity mapping learning in the traditional BP neural network into the learning of F ₂ (x) approximating H(x) in the residual module, and then convert the learning into Learn the residual F ₁ (x). At the same time, in each residual network, the input is connected across one layer of network and output to the next layer of network, so that information can be better transmitted between networks.

Combined with the processing steps of the above-mentioned residual neural network for the input data x, when evaluating the supplier according to the trained supplier evaluation model, the supplier's evaluation parameters are used as the data x in the above steps, and the supplier's evaluation parameters are input In the trained supplier evaluation model, the output of the trained supplier evaluation model, that is, the supplier evaluation result, is obtained through the above-mentioned processing steps of the residual neural network on the input data x. The parameters in the trained supplier evaluation model are continuously optimized during the model training process, and the parameters are finally determined.

FIG. 3 is a flow chart of calculating an evaluation result by a supplier evaluation model according to an embodiment of the present invention. As shown in FIG. 3 , the evaluation parameters of the above-mentioned step S102 are input into the trained supplier evaluation model to obtain the evaluation of the supplier. As a result, steps S301 to S303 may be specifically included.

Step S301, the first-layer network of the supplier evaluation model weights and biases the evaluation parameters through the first-layer network weight and the first-layer network bias, and the result is input to the first-layer network in the first-layer network. In an activation function, the output of the first layer network is obtained.

In this embodiment of the present invention, the first activation function may be a ReLU activation function. The first layer network weights, the first layer network bias and the first activation function of the supplier evaluation model are determined during the training process of the above supplier evaluation model.

Step S302, the second-layer network of the supplier evaluation model weights and biases the output of the first-layer network through the second-layer network weight and the second-layer network offset, and obtains the second-layer network. output.

In the embodiment of the present invention, the second-layer network weight and the second-layer network bias are determined during the training process of the above-mentioned supplier evaluation model.

Step S303, superimpose the output of the second-layer network with the evaluation parameter, and input the superimposed result into the second activation function to obtain the evaluation result of the supplier.

In an optional embodiment of the present invention, the second activation function may be the same as the first activation function. The second activation function is determined during the training of the supplier evaluation model described above.

It can be seen from the above examples that the present invention builds a relatively complete evaluation system, including 6 first-level evaluation parameters and 17 second-level evaluation parameters. The system covers almost all evaluation dimensions, and can accurately and comprehensively evaluate suppliers. Evaluation makes the selection of suppliers more objective. The invention introduces the whole design process of the residual neural network in detail, and establishes a supplier evaluation model based on the residual neural network model. The residual neural network model has a significantly higher matching degree for selecting the best supplier than the artificial neural network model, which overcomes the weakness of the BP neural network gradient disappearance or gradient explosion, and with the increase of the sample set data volume, the accuracy rate is further improved. improve. Therefore, the residual neural network designed in the present invention has more advantages in the field of supplier evaluation.

It should be noted that the steps shown in the flowcharts of the accompanying drawings may be executed in a computer system, such as a set of computer-executable instructions, and, although a logical sequence is shown in the flowcharts, in some cases, Steps shown or described may be performed in an order different from that herein.

Based on the same inventive concept, an embodiment of the present invention also provides a residual neural network-based supplier selection device, which can be used to implement the residual neural network-based supplier selection method described in the above embodiments, as follows example. Since the principle of the residual neural network-based supplier selection device to solve the problem is similar to the residual neural network-based supplier selection method, the embodiment of the residual neural network-based supplier selection device can be found in the residual neural network-based supplier selection method. The embodiment of the supplier selection method will not be repeated here. As used below, the term "unit" or "module" may be a combination of software and/or hardware that implements a predetermined function. Although the apparatus described in the following embodiments is preferably implemented in software, implementations in hardware, or a combination of software and hardware, are also possible and contemplated.

5 is a first structural block diagram of the apparatus for selecting suppliers based on residual neural networks according to an embodiment of the present invention. As shown in FIG. 5 , the apparatus for selecting suppliers based on residual neural networks according to an embodiment of the present invention includes: an evaluation parameter obtaining unit 1 and vendor selection unit 2.

The evaluation parameter obtaining unit 1 is configured to obtain the collected preset evaluation parameters of the supplier.

Supplier selection unit 2, configured to input the evaluation parameters into the trained supplier evaluation model to obtain the evaluation results of the suppliers, and then select suppliers according to the evaluation results of multiple suppliers, wherein the The supplier evaluation model is obtained by using the evaluation parameters marked with the evaluation results as training data and using residual neural network training.

6 is a second structural block diagram of the apparatus for selecting suppliers based on residual neural networks according to an embodiment of the present invention. As shown in FIG. 6 , the apparatus for selecting suppliers based on residual neural networks according to an embodiment of the present invention further includes: a training data acquisition unit 3 and model training unit 4.

The training data obtaining unit 3 is configured to obtain training data, wherein the training data includes: evaluation parameters used for model training and evaluation results corresponding to the marked evaluation parameters used for model training.

The model training unit 4 is used to continuously perform model training according to the training data and the preset residual neural network to obtain the supplier evaluation model, wherein, during model training, the evaluation used for model training is used. The parameters are input into the residual neural network, and the parameters of the residual neural network are adjusted so that the output of the residual neural network approximates the evaluation results corresponding to the evaluation parameters used for model training.

In an optional embodiment of the present invention, the above-mentioned supplier selection unit 2 specifically includes: a first-layer network, a second-layer network, and a processing module.

The first-layer network is used for weighting and biasing the evaluation parameters through the first-layer network weight and the first-layer network bias, and inputting the result into the first activation function to obtain the first-layer output of the network.

The second-layer network is configured to weight and bias the output of the first-layer network by using the second-layer network weight and the second-layer network bias to obtain the output of the second-layer network.

The processing module is configured to superimpose the output of the second layer network and the evaluation parameter, and input the superimposed result into the second activation function to obtain the evaluation result of the supplier.

In order to achieve the above object, according to another aspect of the present application, a computer device is also provided. As shown in FIG. 7 , the computer device includes a memory, a processor, a communication interface and a communication bus, and a computer program that can be run on the processor is stored in the memory, and the processor implements the above embodiments when executing the computer program steps in the method.

The processor may be a central processing unit (Central Processing Unit, CPU). The processor may also be other general-purpose processors, digital signal processors (Digital Signal Processor, DSP), application specific integrated circuit (Application Specific Integrated Circuit, ASIC), Field-Programmable Gate Array (Field-Programmable Gate Array, FPGA) or other Chips such as programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, or a combination of the above types of chips.

As a non-transitory computer-readable storage medium, the memory can be used to store non-transitory software programs, non-transitory computer-executable programs, and units, such as program units corresponding to the above method embodiments of the present invention. The processor executes various functional applications of the processor and works data processing by running the non-transitory software programs, instructions and modules stored in the memory, that is, to implement the methods in the above method embodiments.

The memory may include a storage program area and a storage data area, wherein the storage program area may store an operating system and an application program required by at least one function; the storage data area may store data created by the processor, and the like. Additionally, the memory may include high-speed random access memory, and may also include non-transitory memory, such as at least one magnetic disk storage device, flash memory device, or other non-transitory solid state storage device. In some embodiments, the memory may optionally include memory located remotely from the processor, such remote memory being connectable to the processor via a network. Examples of such networks include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

The one or more units are stored in the memory, and when executed by the processor, perform the methods in the above-described embodiments.

The specific details of the above computer equipment can be understood by referring to the corresponding related descriptions and effects in the above embodiments, which will not be repeated here.

In order to achieve the above object, according to another aspect of the present application, a computer-readable storage medium is also provided, where the computer-readable storage medium stores a computer program, and when the computer program is executed in a computer processor, the above-mentioned based on Steps in a vendor selection method for residual neural networks. Those skilled in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program, and the program can be stored in a computer-readable storage medium. During execution, the processes of the embodiments of the above-mentioned methods may be included. Wherein, the storage medium can be a magnetic disk, an optical disk, a read-only memory (Read-Only Memory, ROM), a random access memory (Random Access Memory, RAM), a flash memory (Flash Memory), a hard disk (Hard Disk Drive) , abbreviation: HDD) or solid-state drive (Solid-State Drive, SSD), etc.; the storage medium may also include a combination of the above-mentioned types of memories.

Obviously, those skilled in the art should understand that the above-mentioned modules or steps of the present invention can be implemented by a general-purpose computing device, which can be centralized on a single computing device, or distributed in a network composed of multiple computing devices Alternatively, they can be implemented with program codes executable by a computing device, so that they can be stored in a storage device and executed by the computing device, or they can be made into individual integrated circuit modules, or they can be integrated into The multiple modules or steps are fabricated into a single integrated circuit module. As such, the present invention is not limited to any particular combination of hardware and software.

The above descriptions are only preferred embodiments of the present invention, and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and changes. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims (12)

1. a kind of supplier selection method based on residual neural network, is characterized in that, comprises: Obtain the preset evaluation parameters of the collected suppliers; Input the evaluation parameters into the trained supplier evaluation model to obtain the evaluation results of the suppliers, so as to select suppliers according to the evaluation results of multiple suppliers, wherein the supplier evaluation model is marked with a good evaluation The evaluation parameters of the results are used as training data and obtained by using residual neural network training. 2. The supplier selection method based on residual neural network according to claim 1, is characterized in that, also comprises: Acquiring training data, wherein the training data includes: evaluation parameters used for model training and evaluation results corresponding to the marked evaluation parameters used for model training; The supplier evaluation model is obtained by continuously performing model training according to the training data and the preset residual neural network, wherein, during model training, the evaluation parameters used for model training are input into the residual In the neural network, the parameters of the residual neural network are adjusted so that the output of the residual neural network approximates the evaluation result corresponding to the evaluation parameters used for model training. 3. the supplier selection method based on residual neural network according to claim 1, is characterized in that, described evaluation parameter is inputted in the supplier evaluation model trained to obtain the evaluation result of described supplier, Specifically include: The first layer network of the supplier evaluation model weights and biases the evaluation parameters through the first layer network weight and the first layer network bias, and the result is input to the first activation function in the first layer network , the output of the first layer network is obtained; The second-layer network of the supplier evaluation model weights and biases the output of the first-layer network through the second-layer network weight and the second-layer network bias, so as to obtain the output of the second-layer network; The output of the second-layer network is superimposed with the evaluation parameters, and the superimposed result is input into the second activation function to obtain the evaluation result of the supplier. 4. The residual neural network-based supplier selection method according to claim 1, wherein the evaluation parameters include: quality level parameters, technical level parameters, supply capacity parameters, procurement cost parameters, and financial risk level parameters and at least one of service level parameters. 5. The residual neural network-based supplier selection method according to claim 4, wherein the quality level parameter comprises: at least one of product qualification rate, excellent batch rate and quality assurance capability parameter; The technical level parameters include: at least one of product research and development capability parameters, product design capability parameters, and independent production capability parameters; The supply capacity parameter includes at least one of: on-time delivery rate, logistics efficiency, supply flexibility, and procurement cost; The procurement cost parameter includes: at least one of a price level parameter, a price rationality parameter, an after-sale guarantee cost, a maintenance cost, and a price reduction capability parameter; The financial risk level parameter includes: whether there is a financial risk indicator; The service level parameters include: response to customer demand rate. 6. A device for selecting suppliers based on residual neural network, characterized in that, comprising: an evaluation parameter obtaining unit, used for obtaining the collected preset evaluation parameters of the supplier; A supplier selection unit, configured to input the evaluation parameters into the trained supplier evaluation model to obtain the evaluation result of the supplier, and then select suppliers according to the evaluation results of multiple suppliers, wherein the supplier The quotient evaluation model is obtained by using the evaluation parameters marked with the evaluation results as training data and using residual neural network training. 7. The residual neural network-based supplier selection device according to claim 6, further comprising: a training data acquisition unit, configured to acquire training data, wherein the training data includes: evaluation parameters used for model training and evaluation results corresponding to the marked evaluation parameters used for model training; A model training unit, configured to continuously perform model training according to the training data and the preset residual neural network to obtain the supplier evaluation model, wherein, during model training, the evaluation parameters used for model training are used for model training. Input into the residual neural network, and adjust the parameters of the residual neural network so that the output of the residual neural network approximates the evaluation result corresponding to the evaluation parameter used for model training. 8. The residual neural network-based supplier selection device according to claim 6, wherein the supplier selection unit specifically comprises: a first-layer network, a second-layer network and a processing module; The first-layer network is used for weighting and biasing the evaluation parameters through the first-layer network weight and the first-layer network bias, and inputting the result into the first activation function to obtain the first-layer the output of the network; The second-layer network is used for weighting and biasing the output of the first-layer network through the second-layer network weight and the second-layer network bias, so as to obtain the output of the second-layer network; The processing module is configured to superimpose the output of the second layer network and the evaluation parameter, and input the superimposed result into the second activation function to obtain the evaluation result of the supplier. 9. The residual neural network-based supplier selection device according to claim 6, wherein the evaluation parameters include: quality level parameters, technical level parameters, supply capacity parameters, procurement cost parameters, and financial risk level parameters and at least one of service level parameters. 10 . The residual neural network-based supplier selection device according to claim 9 , wherein the quality level parameter comprises: at least one of product qualification rate, excellent batch rate and quality assurance capability parameter; 10 . The technical level parameters include: at least one of product research and development capability parameters, product design capability parameters, and independent production capability parameters; The supply capacity parameter includes at least one of: on-time delivery rate, logistics efficiency, supply flexibility, and procurement cost; The procurement cost parameter includes: at least one of a price level parameter, a price rationality parameter, an after-sale guarantee cost, a maintenance cost, and a price reduction capability parameter; The financial risk level parameter includes: whether there is a financial risk indicator; The service level parameters include: response to customer demand rate. 11. A computer device comprising a memory, a processor and a computer program stored on the memory and running on the processor, wherein the processor implements any of claims 1 to 5 when the processor executes the computer program method described in item. 12. A computer-readable storage medium storing a computer program, characterized in that, when the computer program is executed in a computer processor, the computer program according to any one of claims 1 to 5 is implemented. method.

Images