Micro-segmentation-Ruihe Data Technology Holdings Limited-

Micro-segmentation of Customers Based on Deep Learning

Picture

Learning

Features

Accuracy

Overview

Program Overview

Ruihedata solution mines deep features of customers’ asset preference & makes legacy bank models more accurate

It measures the likelihood of customer profile changing states, forecasts how customers may reallocate assets, and gives marketing guidance to bank managers that covers all customer segments.

Contact Demo

Advantages

Solution Advantages

Data Visualization

Data visualization offers an innovative approach to create synergy between banks’ most important business data and deep learning algorithms. It is based on statistics, the law of universal gravitation, space prediction, smooth interpolation and other algorithms.

Autoencoder

“Autoencoder” can extract distinctive features from images, and cluster images with similar features for micro-segmentation.

Bin Coding

The solution applies “bucketing coding” and robust clustering algorithm to ensure precision and substantially improve processing efficiency (up by 400 times).

Values

Solution Value

Hit rate increased by 20%~40%

With image features newly detected by customer micro-segmentation, hit rate of the top 10 percentile in customer list (for NCD and structured deposit) by models increases by 20%-40%.

The hit rate of the top 5% roster has increased by up to 3/4

Model performance strengthens regardless the algorithm applied. In the best case, hit rate of the top 5 percentile in customer list rises by 3/4.

Marketing benefits reach millions to tens of millions

Generating tens of millions of RMB Yuan in direct benefits for marketing.