Identification method for potential customers of auto industry based on tri-training

Abstract

The invention discloses an identification method for potential customers of an auto industry based on tri-training. The method comprises the following steps of data preprocessing and data modeling, wherein the data preprocessing comprises abnormal value processing, missing value processing, classification attribute processing, combination feature generation, feature selection and data normalization; the data modeling adopts a tri-training cooperative training algorithm in semi-supervised learning and selects a BP neural network as a base learner in the process of cooperative training. The historical selling data (customer data with category labels) of an automobile brand dealer and the collected potential customer data (customer data without labels) are utilized to construct a semi-supervised customer identification model, so that accurate marketing targets are provided for accurate marketing of the auto industry, the manpower cost and cost of funds are saved.

Description

technical field [0001] The invention relates to a method for identifying potential customers, in particular to a method for identifying potential customers in the automotive industry based on tri-training. Background technique [0002] With the vigorous development and universal access of the Internet, people's behaviors on the Internet have generated a large amount of data. The demand for storage, processing, and analysis of these data has driven the development of databases, cloud computing and other related technologies. These data collection, storage The development of processing technology has in turn promoted the analysis and application of massive data by enterprises, and promoted the development of big data. At present, more and more enterprises are beginning to pay attention to the accumulation, application and realization of data assets. [0003] In the automobile industry, dealers of various automobile brands can obtain clues of potential consumers who are concer...

AI Technical Summary

Problems solved by technology

[0007] Method 1: It is necessary to designate an experienced salesperson to score each piece of data. Usually, the amount of data of potential customers is very large and is generated every day. Although manual recognition is more accurate than machines to a certain extent, artificial The speed of recognition is very slow, and it may only take a few minutes to use the model to make predictions, while manual recognition may take a day or even a few days, so this method is inefficient and increases labor costs

[0008] Method 2: Hand over the potential customer data to a third-party data company for identification. Since the third-party data companies in the market are uneven, the accuracy of the identification results is difficult to guarantee. At the same time, this kind of customer identification is a long-term demand, so In the long run, the capital cost of seeking a third party is high, and it is difficult to ensure that the third party can fully abide by the data confidentiality treaty, so there is a risk of data leakage

[0009] Method 3: Use a commonly used supervised model to learn the model from the customer data of the purchased car, but in practice, there is a small amount of labeled data, so the model trained with a small amount of data cannot fully grasp the inherent laws of data generation. Therefore, the generalization ability of the model is poor, that is, the prediction accuracy for new data is not high, and it also wastes a lot of information that exists in unlabeled data.

Images