Analysis And Application Of Online-Shopping Data Based On Regional Features

In recent years, on-line trading, as the main form of e-commerce, has gained rapid development in China. With the rise of the juggernaut companies such as Alibaba Group and Jingdong Mall, the development of e-commerce field in China has gone through the platform construction period. E-commerce is now entering the rapid development stage of the medium and small sized e-commerce enterprises which competing with each other on the platforms provided by the big companies. By promoting personalized products and services, these third-party e-commerce enterprises can meet the increasing online trading demands of the public better and achieve the subdivision of e-commerce market and service optimization. However, due to limits of technologies and resources, small and medium-sized e-commerce enterprises have not paid enough attention and application to a large number of transaction data and resources obtained. On one hand, public platforms shall be based on to achieve partial services of online transactions so as to let platform e-commerce enterprises control the user data of core significance for the marketing promotion. On the other hand, due to the lack of applicable analytical methods and technical tools for their sales data, and the insufficient exploration and development for analyzing and exploring their own data models to achieve the personalized services, such enterprises cannot maintain the dominant position in the intense competitions and continuously increase the attraction to potential customers.Generally, small and medium-sized e-commerce enterprises have the features such as regional of services and specialization of products. Thus, on the process of products marketing and service customization, they must pay enough attention to the difference of regional features so as to seek compatible sales forecasting and service recommendation based on different geographical characteristics.Based on the actual sales data of Jiangsu Lianyungang Tianma Network Development Co., Ltd., this thesis aims to build the e-commerce sales forecasting and recommendation models based on the company's sales date combining with the regional features of the buyers. This will help the company in management and operation by analyzing internal and external factors influencing enterprise sales and services. And the main research work of the thesis includes the following contents:(1) Based on its own features of the company in online trading and by cleaning, attribute selection and clustering analysis of e-commerce sales data, this thesis first tries to discover the general commodity classification of regional features such as the seasonality and economic development level. We analyze and discover the most important external factors and the degree of correlation influencing the online sales volume, this work provides reliable base for the region-oriented sales forecasting and service recommendation that following.(2) Based on filtering of regional-feature information influencing the e-commerce sales, we adopt time-series analysis method to build the sales forecasting model based on the regional-feature information and historical sales data so as to propose a forecasting method based on the Hidden Markov model. The method can effectively improve the simple forecasting results based on past sales data and experimentally verify the effectiveness of methods in the thesis.(3) Based on the above general rules of the product sales data and regions features, this thesis purposes a regional oriented product recommendation model based on the genetic algorithm. To improve actual application availability of such models, we further optimized the generalization ability of models and confinement ability of rules.