Predict Shopping Trips (E-Shopping and offline Shopping) based on Deep Learning Approach

Travel production in transportation management has led to extensive studies for different travel purposes. Trips due to the daily activities of citizens are divided with the purposes of work, shopping, education, etc., according to the studies conducted in Tehran, 12% of trips are shopping purposes, which after the trip with the purpose of work. Therefore, by identifying factors affecting the creation of shopping trips, we can play a significant role in reducing transportation cost, reducing pollutant emissions, increasing user satisfaction, and contributing to the growth of sustainable development. In this research, considering the frequency of data in areas 2 and 5 of Tehran metropolis and calculations based on Cochran's formula, 1500 questionnaires were provided to the people of these areas. Finally, 1,000 questionnaires were collected from users in the field of e-commerce who had successful orders in online and offline services in the last 20 days of 2021. The results of the descriptive statistics of the interviewees showed that the largest share of people in the statistical population were single men in the age range of 18-35 years without owning a car and having a bachelor's degree with an income level of 10-15. Also in this paper we used the indicators of age, gender, marital status, car ownership, shopping cost, shopping time, product price, household income, employment status, and education level. It has been used as indicators affecting the type of shopping trip. In the next step, the results were evaluated and the travel type estimated. In order to compare the proposed method, MLP neural network, DT and KNN algorithms were used. The results showed that the deep model has the best performance with an accuracy of 95.63%. After that, there are neural network with 90.12% accuracy, DT with 86.49% accuracy and KNN with 80.16% accuracy.