| dc.description.abstract | With the increase in the number of shoppers on e-commerce websites around the world, companies around the world are trying to concentrate resources on e-commerce websites to improve their competitive advantages. In Taiwan, the handmade soap industry is booming like mushrooms after a spring rain, creating hundreds of millions of business opportunities every year. . Therefore, in the face of such huge business opportunities, the market competition for raw materials for handmade soaps is also very fierce. The research case has been a brick and mortar store with good performance for many years. It is necessary to use the Internet to cross regions and expand market contacts in order to grasp the unlimited and imminent business opportunities of the Internet. In the case study, the main resources of marketing are placed in the e-commerce market, and the business opportunities of operating the website are the development of new website members and the repurchase of old members as the two main forces. In the market where there are multiple options for purchasing manufacturers, customers can change their purchasing behavior at any time, so the marketing community believes that the cost of acquiring a new customer is 5-6 times the cost of retaining an old customer. Therefore, increasing sales resources to explore more business opportunities in the customer base that once purchased, customers continue to return to buy products, and have more and more loyal member customers, which will make our sales work easier and performance. stable growth. Through the literature research, the researches listed include: customer relationship management, website and product visits, click traffic, relationship length with customers, number of purchases, purchase amount, Electronic Direct Mail (EDM) marketing receipts, and openings. , purchase history, preference for special products, return factors, prosperity and unemployment, etc., are all considered as factors related to website behavior related to e-commerce repurchase. Through machine learning techniques, especially random forests, logis regression, and gradient boosting machines to establish predictive repurchase models, the performance of AUC is more than 0.7. In addition, according to the information gain ratio technology provided by Orange, this study judges the importance of different research variables to the target variable according to the information value provided by the target variable, and the top five are ranked as 1. Special products Number of purchases 2. Number of emails sent and opened 3. Number of product purchases after joining as a member 4. Number of times a member has clicked on the homepage of the website in the last three months 5. Number of products cancelled after a member placed an order. Through this research, we grasp the best website behavior variables of customers, devise a product marketing plan to increase the repurchase rate and a strategy to increase the number of members who repurchase, achieve the purpose of constructing a prediction model, effectively attract consumers to repurchase, and continuously create good results. | en_US |