Consumer Behavior Analysis Based on Convolutional Neural Networks

This paper explores the application of convolutional neural systems in the investigation of customer behavior. Due to the ubiquity of the Web and versatile gadgets, companies can gather an expansive sum of shopper behavior information, counting social media intelligent, e-commerce clicks and obtaining behaviors, item surveys, etc. Conventional investigation strategies have restrictions in preparing such complex information, while CNN is able to automatically extract informative features, fitting for training unstructured data like images, videos, and text. To begin with, the essential principles of CNN, its composition, are introduced, including input layer, convolutional layer, activation layer, pooling layer, and totally connected layer, along with their functions in feature extraction and model establishment. After that, it goes to demonstrate the pre-processing and visualization of a retail sales and customer behavior dataset and constructs a CNN model to predict a classification of whether a customer is churned or not. It demonstrates very perfect performances in terms of accuracy, AUC, precision, and F1 score from the CNN model. Moreover, the paper compares CNN against a number of alternative classification methods: logistic regression, support vector machine, K-nearest neighbor, decision tree, random forest, plain Bayes, XGBoost, and multilayer perceptron. Comparative results demonstrate that CNN models work better in most classic classifiers on consumer behavior analysis. The article finally summarizes the potential of CNNs in consumer behavior analysis and looks forward to the future with the continuous progress of deep learning technology and rich data sources. It will help enterprises step into a digital transformation and make decisions intelligently.