The application of deep learning algorithm in marketing intelligence

58ganji company had accumulated a large number of users, sales staffs by virtue of their experience in which to select target customers and carry out sales work had become inadequate. By using deep learning model, potential customers are scored through their historical behavior data, and their posterior probability value of transaction is given, after sorted from largest to smallest, pushed to sales man. This method provides accurate target customers for marketing work to achieve the goal of saving amount of labor for salesperson to select high quality customer and effectively improving the close rate, so as to achieve the goal of bringing greater economic benefits to company. In this paper, Lightgbm and LR combination models were used for comparison, the AUC scores in the six validation sets by using deep learning model VS Lightgbm+LR model were 0.82, 0.78 for group 1, 0.88, 0.81 for group 2, 0.92, 0.86 for group 3, 0.93, 0.88 for group 4, 0.95, 0.92 for group 5, and 0.97, 0.95 for group 6. Nearly 80,000 eligible online users were predicted in the latest month, there were 344 customers became members, Among the top 20,000 users using machine learning model and deep learning model scoring ranking, the number of transactions is 183 and 225, The forecast coverage is 53% and 67.3% respectively. The results show that the deep learning model performs better in precision marketing.