C4.5

Abstract: Data Mining is a very interesting area to mine the data for knowledge. Several techniques are available. A data mining is one of the fast growing research field which is used in a wide areas of applications. The data mining consists of classification algorithms, association algorithms and searching algorithms. We are using clustering and decision tree methods to mine the data by using hybrid algorithms K-means, SOM and HAC algorithms from clustering and ID3 and C4.5 and CART algorithms from decision tree and it can produce the better results than the traditional algorithms. The comparative study of these algorithms to obtain which one is high accuracy, frequency, measure, procedure, Pruning, error rate and time complexities from the decision tree algorithm. At the same time the comparative study of these algorithms to obtain which one is quality, error rate, computation time, accessing time, performance, number of cluster, Map topology, type of software, type of data, and data set size from the clustering algorithm.
Key words: K-MEANS, SOM, HAC, ID3, C4.5, CART
Title: Comparative Study of different clustering and Decision Tree for Data Mining Algorithm
Author: Ms. A. Sivasankari,  Mrs. S. Sudarvizhi, S. Radhika Amirtha Bai
International Journal of Computer Science and Information Technology Research
ISSN 2348-120X (online), ISSN 2348-1196 (print)
Research Publish Journals

Classification is one of the techniques contained in data mining. The purpose or objective of data mining classification technique is to predict accurately target class using variables related. There are many models of algorithms in data mining classification techniques. Klasfikasi algorithm model has a different value and highly dependent on the number of attributes and records of the dataset.
The selected dataset relates to readmission of diabetic patients. The dataset still contains missing values so that in this study, the data prepocessing stage is developed. After prepocessing phase, the selected dataset consists of 47 attributes and 49 735 records. In this study also, Decision Tree algorithm performance will be explored by the selected dataset. The Decision Tree Classifiers which are evaluated  are ID3, C4.5 and CART. 10-fold Cross Validation technique was developed to validate or generate performance result of each classifier.
This study shows that C4.5 classifier outperforms ID3 and CART.  C4.5 results 54.13% of  accuracy  and  6 seconds execution time performance.

ان تعدين البيانات فعالية الحصول على المعرفة لتحقيق هدف أساس وهو اكتشاف الحقائق الخفية (Hidden Facts) التي تتضمنها قواعد البيانات وذلك من خلال استخدام تقنيات متعددة تشتمل على الذكاء الاصطناعي، التحليلات الاحصائية، تقنيات و نمذجة البيانات ... الخ. فأن عملية تعدين البيانات تولد نماذج وعلاقات واضحة في البيانات والتي تساعد على توقع النتائج في المستقبل. وقد ظهرت العديد من الخوارزميات التي في هذا المجال، وترتب عليها مقارنة بين هذه الخوارزميات لاختبار الخوارزمية المناسبة في الحصول على نتائج أفضل. وقد هدف البحث الى استخدام المصنف C4.5 وربطها مع الشبكة العصبية نوع Back –  Propagation (BP) وذلك لتكوين نموذج تصنيف يحمل خواص الطرفين، فضلا عن مقارنة النتائج المستحصلة مع نتائج التصنيف باستخدام الحزمة البرمجية الجاهزة Minitab . وتوصل البحث الى ان المعادلات الخاصة بالمصنف C4.5 كانت أكفأ في الاداء وخاصة بعد ربطها بالشبكة العصبية BP لازالة التناقض والتشويش الموجود في البيانات، كما عززت النتائج من افضلية استخدام لغات البرمجة مقارنة بنتائج التطبيق الجاهزة.

RoboCup is an international research aimed at improving artificial intelligence and robotics. It is a standard issue to get a wide range of technologies together and obtain new achievements. In 2D simulation league, after each game, server saves a log that contains all information about the game. By using data mining techniques knowledge can be discovered from this massive data. In this research we aimed to extract ball and player positions from log files and pre-process this data to specify some information including the action that have been taken place, start point, and involving players, etc. We mined this data to predict own and opponent action using C4.5 algorithm. The result showed that after applying our method the goal scoring was increased 251.39% with 64.13% confidence interval (with alpha = 0.1).

Mobile adhoc networks, abbreviated as MANETs are more vulnerable to various attacks than other types of networks. This is because of natural properties of MANETs such as dynamic topology, ever changing infrastructure and lack of central controller. Hence detecting anomaly behaviours in MANETs is a challenging task and this topic has attracted many researchers. In this paper, clustering method based anomaly detection method is proposed. The method uses C4.5 clustering algorithm to classify the events as normal or abnormal. The proposed method works in three phases viz. training, detection of anomaly behaviour and identifying the type of attack.
The proposed method uses 141 features of network to identify the attacks. The experiments were conducted on NS2 simulator and results show that the proposed method is effective in detecting anomalies. Moreover, False Alarm Ratio (FAR) is also low which ensure that the detected anomalies are real attacks.

The aim of data mining is to extract hidden knowledge from huge amount of data set and generate clear and easy understandable patterns. Diabetes is a group of metabolic disease caused by increased level of blood glucose. Different data mining algorithms are applied in medical research in order to diagnosis large amount of medical dataset. Various data mining algorithms were designed for diagnosing diabetes based on physical and chemical tests. The main data mining algorithms discussed in this paper are EM algorithm, K means, C4.5 algorithm, Genetic algorithm and SVM. EM is the expectation maximization used for sampling, to determine and maximize the expectation in successive iterative cycles. C4.5 is a decision tree induction technique that has been successfully applied for medical data. Genetic algorithm is population based model that uses selection and recombination operators to generate new sample points. Support Vector Machine are set of supervised learning method whose training tech permit to represent complex non linear function. K means is a unsupervised which objects are moved among sets of cluster until the desired set is reached. This paper studies the comparison of various data mining algorithms for prediction of diabetes disease.

With the promises of predictive analytics in big data, and the use of machine learning algorithms, predicting future is no longer a difficult task, especially for health sector, that has witnessed a great evolution following the development of new computer technologies that gave birth to multiple fields of research. Many efforts are done to cope with medical data explosion on one hand, and to obtain useful knowledge from it, predict diseases and anticipate the cure on the other hand. This prompted researchers to apply all the technical innovations like big data analytics, predictive analytics, machine learning and learning algorithms in order to extract useful knowledge and help in making decisions. In this paper, we will present an overview on the evolution of big data in healthcare system, and we will apply three learning algorithms on a set of medical data. The objective of this research work is to predict kidney disease by using multiple machine learning algorithms that are Support Vector Machine (SVM), Decision Tree (C4.5), and Bayesian Network (BN), and chose the most efficient one.

Pemanfaatan data mining dalam strategi pemasaran perbankan sangat efektif. Segmentasi nasabah pensiunan yang akan ditawarkan produk kredit merupakan salah satu proses yang dilakukan dalam strategi pemasaran perbankan. Untuk mendukung hasil dari tingkat keberhasilan tenaga pemasaran dalam perannya untuk memasarkan produk kredit perbankan yang prosesnya membutuhkan data-data nasabah pensiunan ini, maka dukungan data mining sangat berperan penting dalam klasifikasi calon nasabah bank sehingga dapat memprediksi tingkat keberhasilan dalam pemasaran produk layanan tersebut. Berdasarkan pemetaan penelitian mengenai dukungan data mining pada nasabah pensiunan maka penelitian ini diputuskan untuk membanding 2 metode klasifikasi data mining, yaitu Naive Bayes dan C4.5

In this article, we evaluate features and algorithms for the task of prosodic boundary prediction for Greek. For this purpose a prosodic corpus composed of generic domain text was constructed. Feature contribution was evaluated and ranked with the application of information gain ranking and correlation-based feature selection filtering methods. Resulted datasets were applied to C4.5 decision tree, one-neighbour instance based learner and Bayesian learning methods. Models performance exploitation led as to the construction of a practically optimal feature set whose prediction effectiveness was evaluated with two prosodic databases. In terms of total accuracy and F-measure, evaluation results established the decision tree effectiveness in learning rules for prosodic boundary prediction.