Research on segmentation and classification of bank customer intention recognition accuracy based on large model

The performance of various machine learning models for customer segmentation tasks is explored around neural networks, decision trees and support vector machines (SVMS) applied to standardized data sets. We collect customer data and train classification models using these collection labels. The evaluation is based on accuracy - leakage rate F. The neural network obtains the best result of normalized data with the ROC-AUC of 0.986, 0.971 and 0.998. Its accuracy drops to 0.944 on non-normalized data. Decision trees have stable performance on standardized data sets with ROC-AUC scores of 0.954 and 0.951, respectively. Support vector machines performed well on standardized data with ROC-AUC scores of 0.940 and 0.990 but poorly on non-standardized data especially on neural networks and support vector machines. This study highlights the important impact of data standardization, providing valuable guidance for future applications in customer segmentation and ultimately refining customer retention marketing strategies with an in-depth understanding of machine learning curves and the performance of these models.