Development of Prediction Methods for Taxi Order Service on the Basis of Intellectual Data Analysis

Abstract

The work considers the urgent task of collecting and analyzing information received during the work of the taxi order service. The data obtained by the taxi service can be easily represented by different time series. Particular attention is also paid to the use of neural networks to solve the predicting problem. The relevance of using neural networks in comparison with statistical models is substantiated. The special software used allows one’s to collect information on the operation of the service in a variety of SQL tables. Particular attention is paid to existing programming languages that allow to implement data mining processes. The strengths and weaknesses are highlighted for this languages. Based on the accumulated data on the numbers of taxi service orders, the algorithms for predicting the operation of a taxi service were studied using both neural networks and mathematical models of random processes. Comparative predicting characteristics are obtained, variances of predicting errors are found. The results of construction using autoregressive and doubly stochastic models, as well as using fuzzy logic models, are presented. It is shown that the use of neural networks provides smaller errors in predicting the number of taxi service orders.