CN107679668A - The electric bicycle travel time prediction method of duration model based on risk - Google Patents

Abstract

The invention discloses a method for predicting travel time of an electric bicycle based on a risk duration model. Firstly, obtain the personal and family characteristics and all-day travel information of electric bicycle users; secondly, extract the parameters corresponding to different divisions under different information items; then, divide the total sample into male and female according to gender; then, use The risk duration model builds a prediction model for the travel time of electric bicycles for men and women; predicts the travel time of electric bicycles through the risk duration model, and analyzes the impact of various factors on the travel time of electric bicycles based on the results. The travel time of the electric bicycle can be predicted by adopting the method of the invention, which is helpful for professionals to predict the use demand of the electric bicycle, thereby formulating policies for promoting the use of electric bicycles in cities and making infrastructure planning well.

Description

Risk-Based Duration Model for Electric Bicycle Travel Time Prediction Method

technical field

The invention relates to a method for predicting the travel time of an electric bicycle, in particular to a method for predicting the travel time of an electric bicycle based on a risk duration model.

Background technique

.Compared with cars, electric bicycles are an environmentally friendly and sustainable means of transportation. The use of electric bicycles can help solve urban traffic and pollution problems such as traffic congestion, accident deaths, energy consumption and deterioration of air quality; the utilization rate of electric bicycles The interrelationship between various influencing factors plays an important role in different traffic planning work. Therefore, electric bicycles are an indispensable research topic in the field of transportation in China.

When reviewing previous studies, it was found that most studies did not distinguish between men and women, but in fact, gender is an important factor in determining travel behavior, and women's physical and psychological characteristics make their travel time somewhat different from that of men. In addition, most studies focus on the choice of electric bicycle mode, and few studies consider the time of bicycle travel, which has a significant impact on traffic demand and is an important factor in determining traffic demand. The study of e-bike travel time can provide a better understanding of the influence of various factors on bicycle use and will help transportation professionals predict the demand for e-bike use, which is the basis for developing effective policies to promote e-bike use and An important prerequisite for good infrastructure planning.

Contents of the invention

Technical problem: The present invention provides a method for predicting travel time of electric bicycles based on a risk-based duration model, which can be used to analyze the travel time of electric bicycles, and helps to formulate policies to promote the use of electric bicycles in cities and to improve infrastructure planning.

Technical solution: A method for predicting travel time of an electric bicycle based on a risk-based duration model according to the present invention comprises the following steps:

(1) Obtain the personal and family characteristics of electric bicycle users and their travel information;

(2) Extracting parameters corresponding to different divisions under different information items;

(3) Divide the total sample into male and female according to gender;

(4) A risk-based duration model that models male and female e-bike travel times separately;

(5) Predict the travel time of electric bicycles, and analyze the influence of various factors on the travel time of electric bicycles according to the model results.

The personal and family characteristics and travel information mentioned in the step (1) mainly include: traveler’s occupation, traveler’s age, family annual income, car ownership, travel purpose, travel distance, travel starting point population density, travel duration, The population density of the travel destination, whether the travel time is morning peak, and the traffic flow of the origin and destination.

The parameters in the step (2) are set as follows: the occupation of the traveler is student, worker, official and others, and the corresponding parameters are x _1i , x _2i , x _3i , x _4i ; the age of the traveler is less than 20 years old, 20 to 40 For those aged between 40 and 50 and over 50, the corresponding parameters are x _5i , x _6i , x _7i , x _8i ; for those whose annual family income is less than 2,000 RMB and greater than 20,000 RMB, the corresponding parameters are x _9i , x _10i ; I have a car at home, I will buy a car in the next five years, I will buy a car in the next ten years, and I will not buy a car in the future, the corresponding parameters are x _11i , x _12i , x _13i , x _14i ; the purpose of travel is work, school, shopping , home and others, the corresponding parameters are x _15i , x _16i , x _17i , x _18i , x _19i ; the travel purposes are work, school, shopping, home, and other travel distances, and the corresponding parameters are x _20i , x _21i , x _22i , x _23i , x _24i , x _25i ; the population density of the travel starting point is greater than 0.023 persons/square meter and less than 0.023 persons/square meter, the corresponding parameters are x _26i , x _27i ; the population density of the travel destination is greater than 0.023 persons/square meter sum is less than 0.023 people/square meter, the corresponding parameters are x _28i , x _29i ; the travel time is the morning peak, the corresponding parameter is x _30i ; the traffic flow of the origin and destination points, the corresponding parameter is x _31i ; other information, the corresponding parameter x _ki ; i represents the i-th questionnaire.

Described step (4) comprises the following steps:

(41) Based on the risk-based duration model, the cumulative distribution function F(t) of travel time t is as follows:

where f(t) is the probability density function of the travel time t, S(t) is the survival function which gives the probability that the duration is greater than t, the hazard function h(t) gives the probability of ending the trip at time t, and the condition is that the trip has not ended before time t, the equation is as follows:

(42) Use the accelerated failure time model as the hazard function to explain the influence of explanatory variables in the risk-based duration model. The expression of the accelerated failure time model is as follows:

h(t _i )＝h ₀ (t×exp(-(β ₀ +β ₁ x _1i +…β _n x _ni )))×exp(-(β ₀ +β ₁ x _1i +…β _n x _ni ) )

where h0(.) is the baseline hazard function, n is the number of explanatory variables, and the accelerated failure time model can also be written as the following expression:

ln(t)＝β ₀ +β ₁ x _1i +β ₂ x _2i +β ₃ x _3i +β ₄ x _4i +β ₅ x _5i +β ₆ x _6i +β ₇ x _7i +β ₈ x _8i +β ₉ x _9i +β ₁₀ x _10i +β ₁₁ x _11i +β ₁₂ x _12i +β ₁₃ x _13i +β ₁₄ x _14i +β ₁₅ x _15i +β ₁₆ x _16i +β ₁₇ x _17i +β ₁₈ x _18i +β ₁₉ x _19i +β ₂₀ x _20i +β ₂₁ x _21i +β ₂₂ x _22i +β ₂₃ x _23i +β ₂₄ x _24i +β ₂₅ x _25i +β ₂₆ x _26i +β ₂₇ x _27i +β ₂₈ x _28i +β ₂₉ x _29i +β ₃₀ x _30i +β ₃₁ x _31i +β _k x _ki +θ+σε _i

where t represents the travel time, εi is the independent random error, σ is the scale parameter, θ is the random effect, and _βk is the corresponding _coefficient ;

(43) Use the Weibull distribution to estimate the hazard function. The probability density function and hazard function of the Weibull distribution are as follows:

f(t|λ,p)=λp(λt) ^p-1 e ^-(λt)p ,h(t)=λ ^p pt ^p-1 .

Among them, the introduction of random effects to the accelerated failure time model is used to explain the heterogeneity of the entire sample.

Beneficial effects: Compared with the prior art, the present invention has the beneficial effects: 1. Studying the travel time of electric bicycles can better understand the influence of various factors on the use of electric bicycles, and help transportation professionals predict electric bicycles. The demand for bicycle use is also an important prerequisite for formulating effective policies and good infrastructure planning; 2. Divide the total sample into two categories, male and female, and model them separately to make the prediction more accurate; 3. The present invention uses accelerated failure time model to analyze the travel time of electric bicycles, which can better capture the direct impact of explanatory variables on the travel time of electric bicycles; 4. This method uses Weibull distribution to estimate the risk function, through Weibull, lognormal, normal The likelihood ratio test of the four distributions of normal and exponential distributions is carried out. The results prove that the likelihood function ratio of the Weibull distribution is the largest, and this distribution is most suitable for the travel time data of electric bicycles.

Description of drawings

Fig. 1 is a block flow diagram of the present invention;

Figure 2 is the mean absolute percentage error of the male model;

Figure 3 is the mean absolute percentage error of the female model;

Figure 4 is a male AFT model with random effects;

Figure 5 is the male AFT model without random effects;

Figure 6 is a female AFT model with random effects;

Figure 7 is the female AFT model without random effects.

detailed description

The present invention will be further described below in conjunction with the accompanying drawings of the description.

Obtain personal and family characteristics and travel information of electric bicycle users, mainly including: traveler's occupation, traveler's age, family annual income, car ownership, travel purpose, travel distance, travel starting point population density, travel duration, travel destination population Density, whether the travel time is the morning peak and the traffic flow of the origin and destination.

Extract the parameters corresponding to different divisions under different information items: the traveler’s occupation is student, worker, official and others, and the corresponding parameters are x _1i , x _2i , x _3i , x _4i ; the traveler’s age is less than 20 years old, 20 years old For those between the ages of 40, 40 to 50 and over 50, the corresponding parameters are x _5i , x _6i , x _7i , x _8i ; for those whose annual family income is less than 2,000 RMB and greater than 20,000 RMB, the corresponding parameter is x _9i , x _10i ; I have a car at home, I will buy a car in the next five years, I will buy a car in the next ten years, and I will not buy a car in the future, the corresponding parameters are x _11i , x _12i , x _13i , x _14i ; the purpose of travel is work and school , shopping, going home, and others, the corresponding parameters are x _15i , x _16i , x _17i , x _18i , x _19i ; the travel purposes are work, school, shopping, home, and other travel distances, and the corresponding parameters are x _20i , x _21i , x _22i , x _23i , x _24i , x _25i ; the population density of the travel starting point is greater than 0.023 persons/square meter and less than 0.023 persons/square meter, the corresponding parameters are x _26i , x _27i ; the population density of the travel destination is greater than 0.023 persons/square meter If the sum of square meters is less than 0.023 people/square meter, the corresponding parameters are x _28i and x _29i ; the travel time is morning peak, the corresponding parameter is x _30i ; the traffic flow at the origin and destination points, the corresponding parameter is x _31i ; for other information, the corresponding parameter is x _ki ; i represents the i-th questionnaire.

The total sample was divided into male and female parts according to gender.

A risk-based duration model that models male and female e-bike travel times separately:

Based on the risk-based duration model, the cumulative distribution function F(t) of travel time t is as follows:

where f(t) is the probability density function of the travel time t, S(t) is the survival function which gives the probability that the duration is greater than t, the hazard function h(t) gives the probability of ending the trip at time t, and the condition is that the trip has not ended before time t, the equation is as follows:

The accelerated failure time model (AFT) is used as the hazard function to explain the influence of explanatory variables in the risk-based duration model. The expression of the accelerated failure time model is as follows:

h(t _i )＝h ₀ (t×exp(-(β ₀ +β ₁ x _1i +…β _n x _ni )))×exp(-(β ₀ +β ₁ x _1i +…β _n x _ni ) )

where h0(.) is the baseline hazard function, n is the number of explanatory variables, and the accelerated failure time model can also be written as the following expression:

ln(t)＝β ₀ +β ₁ x _1i +β ₂ x _2i +β ₃ x _3i +β ₄ x _4i +β ₅ x _5i +β ₆ x _6i +β ₇ x _7i +β ₈ x _8i +β ₉ x _9i +β ₁₀ x _10i +β ₁₁ x _11i +β ₁₂ x _12i +β ₁₃ x _13i +β ₁₄ x _14i +β ₁₅ x _15i +β ₁₆ x _16i +β ₁₇ x _17i +β ₁₈ x _18i +β ₁₉ x _19i +β ₂₀ x _20i +β ₂₁ x _21i +β ₂₂ x _22i +β ₂₃ x _23i +β ₂₄ x _24i +β ₂₅ x _25i +β ₂₆ x _26i +β ₂₇ x _27i +β ₂₈ x _28i +β ₂₉ x _29i +β ₃₀ x _30i +β ₃₁ x _31i +β _k x _ki +θ+σε _i

where t represents the travel time, εi is the independent random error, σ is the scale parameter, θ is the random effect, and _βk is the corresponding _coefficient ;

Use the Weibull distribution to estimate the hazard function. The probability density function and hazard function of the Weibull distribution are as follows:

f(t|λ,p)=λp(λt) ^p-1 e ^-(λt)p ,h(t)=λ ^p pt ^p-1 .

Among them, the introduction of random effects to the accelerated failure time model is used to explain the heterogeneity of the entire sample.

Using household travel data from an extensive survey in Shaoxing, China in 2007, Shaoxing is a typical medium-sized city located on the east coast of China with a population of 908,500 and a total area of 59.96 square kilometers in 2007. A total of 7,320 questionnaires were used.

Table 1 shows the results of the electric bicycle travel time prediction model constructed based on the risk-based duration model.

Table 1 Estimation results of the accelerated failure time model with random effects

Note: ^a standard deviation;

^b This variable is not important in the model;

^cReference level.

From the above table, we can draw the following conclusions: the estimated parameters of travel distance for both men and women are positive numbers, indicating that with the increase of travel distance, the travel time of electric bicycles for both male and female travelers may be longer; for different trips Objective, the travel distance parameter in the male model is smaller than the travel distance parameter in the female model. The possible explanation for this is that the travel speed of female travelers is lower than that of male travelers; E-bikes traveled shorter distances than men; moreover, the parameter of working distance was lower than the other parameters of travel distance, suggesting that e-bike riders were more likely to travel faster on commuting trips.

The parameters for origin-destination traffic volume are all positive, indicating that the duration of e-bike trips increases with origin-destination traffic volume; however, the parameters in the male model are larger than those in the female model, a possible explanation for this is that, Traffic congestion has a greater impact on the duration of male e-bike trips, because men may have longer travel distances; higher population density at the origin and destination will lead to a reduction in travel time for both men and women, because the road network between these areas The traffic capacity is large, thereby reducing the driving time of the electric bicycle.

Regarding the age of travel, the parameters show that: men aged 20 to 50 are likely to travel less than men aged 50+, and conversely women aged 20 to 50 are likely to travel longer than women aged 50+ The explanation for this phenomenon may be related to the difference in travel patterns between older men and older women. Older men may prefer long-term travel due to retirement, while older women may prefer short-term travel due to family responsibilities. travel.

Regarding the travel purpose variable, if the travel purpose is work, then men will travel longer than women, and the shopping parameter for men is negative, but this parameter is not important for women, indicating that men spend less time shopping .

The morning rush hour and occupation also affect e-bike travel time, for both men and women, if an e-bike trip is taken during the morning rush hour, the trip time is more likely to be longer, which may be due to traffic congestion during the morning rush hour; men and women Female workers travel longer by e-bike because they are likely to live in the suburbs.

To evaluate the predictive performance of the model, the error relative to the travel time observations was tested using the mean absolute percentage error (MAPE), which is defined as:

where t _A (i) represents the actual value of the i-th observation, and t _p (i) represents the predicted value of the i-th observation. Figures 2 and 3 show the accelerated failure time (AFT) model for different trip durations. MAPE values, the MAPEs of the random effects AFT model were lower than those of the AFT model, and there was no random effect on different durations.

Previous researchers have used MAPE thresholds of 50%, 20% and 10% to represent reasonable, good and high-precision levels to evaluate the predictive performance of risk-based duration models respectively. The random effect AFT model for men and women in Table 1 The overall MAPEs are 10.4% and 11.8%, indicating that the developed AFT model has better prediction accuracy; for comparison, the MAPEs of the AFT model without random effects are also calculated at the same time, and the results are 39.5% and 40.8%, therefore, including Random effects can also improve the predictive performance of the AFT model. Figure 4, Figure 5, Figure 6, and Figure 7 show the actual and predicted value diagrams of the AFT model with and without random effects, which also show that the random effect AFT model can provide better predictive performance.

Claims (5)

1. A method for predicting the travel time of an electric bicycle based on a risk-based duration model, characterized in that, comprising the following steps: (1) Obtain the personal and family characteristics of electric bicycle users and their travel information; (2) Extracting parameters corresponding to different divisions under different information items; (3) Divide the total sample into male and female according to gender; (4) A risk-based duration model that models male and female e-bike travel times separately; (5) Predict the travel time of electric bicycles, and analyze the influence of various factors on the travel time of electric bicycles according to the model results. 2. The electric bicycle travel time prediction method based on the risk-based duration model of claim 1, characterized in that, the personal and family characteristics and travel information mentioned in the step (1) mainly include: traveler's occupation , traveler's age, family annual income, car ownership, travel purpose, travel distance, travel origin population density, travel duration, travel destination population density, whether the travel time is the morning peak and the traffic flow at the origin and destination. 3. the electric bicycle travel time prediction method based on the duration model of risk according to claim 1, is characterized in that, the parameter setting in the described step (2) is: traveler's occupation is student, worker, official and other , and its corresponding parameters are x _1i , x _2i , x _3i , x _4i ; the corresponding parameters are x _5i , x _6i , x _7i , x _8i ; the corresponding parameters are x _9i , x _10i if the annual household income is less than 2,000 RMB and greater than 20,000 RMB; the family has a car, will buy a car in the next five years, will buy a car in the next ten years, and will not buy a car in the future Will buy a car, the corresponding parameters are x _11i , x _12i , x _13i , x _14i ; the purpose of travel is work, school, shopping, home and others, the corresponding parameters are x _15i , x _16i , x _17i , x _18i , x _19i ; Travel purposes are work, school, shopping, home, and other travel distances, and the corresponding parameters are x _20i , x _21i , x _22i , x _23i , x _24i , x _25i ; the population density of the travel starting point is greater than 0.023 people/square meter and less than 0.023 people/square meter, the corresponding parameters are x _26i and x _27i ; the population density of the travel terminal is greater than 0.023 people/square meter and less than 0.023 people/square meter, the corresponding parameters are x _28i and x _29i ; the travel time is the morning peak, The corresponding parameter is x _30i ; the traffic flow at the origin and destination point is x _31i ; other information is the corresponding parameter x _ki ; i represents the i-th questionnaire. 4. the electric bicycle travel time prediction method based on the duration model of risk according to claim 1, is characterized in that, described step (4) comprises the following steps: (41) Based on the risk duration model, the cumulative distribution function F(t) of travel time t is as follows: <mrow><mi>F</mi><mrow><mo>(</mo><mi>t</mi><mo>)</mo></mrow><mo>=</mo><msubsup><mo>&amp;Integral;</mo><mn>0</mn><mi>t</mi></msubsup><mi>f</mi><mrow><mo>(</mo><mi>T</mi><mo>)</mo></mrow><mi>d</mi><mi>T</mi><mo>=</mo><mi>Pr</mi><mrow><mo>(</mo><mi>T</mi><mo>&amp;le;</mo><mi>t</mi><mo>)</mo></mrow><mo>=</mo><mn>1</mn><mo>-</mo><mi>Pr</mi><mrow><mo>(</mo><mi>T</mi><mo>&gt;</mo><mi>t</mi><mo>)</mo></mrow><mo>=</mo><mn>1</mn><mo>-</mo><mi>S</mi><mrow><mo>(</mo><mi>t</mi><mo>)</mo></mrow><mo>,</mo><mn>0</mn><mo>&lt;</mo><mi>x</mi><mo>&lt;</mo><mi>&amp;infin;</mi></mrow> where f(t) is the probability density function of the travel time t, S(t) is the survival function which gives the probability that the duration is greater than t, the hazard function h(t) gives the probability of ending the trip at time t, and the condition is that the trip has not ended before time t, the equation is as follows: <mrow><mi>h</mi><mrow><mo>(</mo><mi>t</mi><mo>)</mo></mrow><mo>=</mo><munder><mi>lim</mi><mrow><mi>&amp;Delta;</mi><mi>t</mi></mrow></munder><mfrac><mrow><mi>Pr</mi><mrow><mo>(</mo><mi>t</mi><mo>&amp;le;</mo><mi>T</mi><mo>&amp;le;</mo><mi>t</mi><mo>+</mo><mi>&amp;Delta;</mi><mi>t</mi><mo>|</mo><mi>T</mi><mo>&amp;GreaterEqual;</mo><mi>t</mi><mo>)</mo></mrow></mrow><mrow><mi>&amp;Delta;</mi><mi>t</mi></mrow></mfrac><mo>=</mo><mfrac><mrow><mi>f</mi><mrow><mo>(</mo><mi>t</mi><mo>)</mo></mrow></mrow><mrow><mn>1</mn><mo>-</mo><mi>F</mi><mrow><mo>(</mo><mi>t</mi><mo>)</mo></mrow></mrow></mfrac><mo>=</mo><mfrac><mrow><mi>f</mi><mrow><mo>(</mo><mi>t</mi><mo>)</mo></mrow></mrow><mrow><mi>S</mi><mrow><mo>(</mo><mi>t</mi><mo>)</mo></mrow></mrow></mfrac><mo>;</mo></mrow> (42) Use the accelerated failure time model as the hazard function to explain the influence of explanatory variables in the risk-based duration model. The expression of the accelerated failure time model is as follows: h(t _i )＝h ₀ (t×exp(-(β ₀ +β ₁ x _1i +…β _n x _ni )))×exp(-(β ₀ +β ₁ x _1i +…β _n x _ni ) ) where h0(.) is the baseline hazard function, n is the number of explanatory variables, and the accelerated failure time model can also be written as the following expression: ln(t)＝β ₀ +β ₁ x _1i +β ₂ x _2i +β ₃ x _3i +β ₄ x _4i +β ₅ x _5i +β ₆ x _6i +β ₇ x _7i +β ₈ x _8i +β ₉ x _9i +β ₁₀ x _10i +β ₁₁ x _11i +β ₁₂ x _12i +β ₁₃ x _13i +β ₁₄ x _14i +β ₁₅ x _15i +β ₁₆ x _16i +β ₁₇ x _17i +β ₁₈ x _18i +β ₁₉ x _19i +β ₂₀ x _20i +β ₂₁ x _21i +β ₂₂ x _22i +β ₂₃ x _23i +β ₂₄ x _24i +β ₂₅ x _25i +β ₂₆ x _26i +β ₂₇ x _27i +β ₂₈ x _28i +β ₂₉ x _29i +β ₃₀ x _30i +β ₃₁ x _31i +β _k x _ki +θ+σε _i where t represents the travel time, εi is the independent random error, σ is the scale parameter, θ is the random effect, and _βk is the corresponding _coefficient ; (43) Use the Weibull distribution to estimate the hazard function. The probability density function and hazard function of the Weibull distribution are as follows: <mrow><mi>f</mi><mrow><mo>(</mo><mi>t</mi><mo>|</mo><mi>&amp;lambda;</mi><mo>,</mo><mi>p</mi><mo>)</mo></mrow><mo>=</mo><mi>&amp;lambda;</mi><mi>p</mi><msup><mrow><mo>(</mo><mi>&amp;lambda;</mi><mi>t</mi><mo>)</mo></mrow><mrow><mi>p</mi><mo>-</mo><mn>1</mn></mrow></msup><msup><mi>e</mi><mrow><mo>-</mo><msup><mrow><mo>(</mo><mi>&amp;lambda;</mi><mi>t</mi><mo>)</mo></mrow><mi>p</mi></msup></mrow></msup><mo>,</mo><mi>h</mi><mrow><mo>(</mo><mi>t</mi><mo>)</mo></mrow><mo>=</mo><msup><mi>&amp;lambda;</mi><mi>p</mi></msup><msup><mi>pt</mi><mrow><mi>p</mi><mo>-</mo><mn>1</mn></mrow></msup><mo>.</mo></mrow> 5. The electric bicycle travel time prediction method based on the risk-based duration model of claim 4, wherein the introduction of the random effect is used to explain the heterogeneity of the entire sample.