Solutions - K-Prototype Segmentation

Accurate customer identification and segmentation

Monitor Plus K-Prototype Segmentation™

is a specialized system for risk factor segmentation. This module is useful for the identification of groups of clients and factors with similar characteristics based on a large number of variables (categorical and numerical).

Use cases

  • Identification, measurement, control and monitoring of segments.
  • Identify and manage specific risks applicable to each segment.
  • Comprehensive understanding of transactional features.
  • Reduction of false positives in the detection of suspicious transactions.

MAIN BENEFITS

In-depth analysis of customer transactional behavior

Monitor Plus K-Prototype Segmentation™ natively integrates three machine learning algorithms (K-Means, K-Modes and K-Prototype) to provide a comprehensive analysis of each customer's transactional behavior.

Segment change detection

The model analyzes variables, generates descriptive statistics, allows for the treatment of outliers and has a periodic monitoring tool that analyzes changes in a customer's segment in the four risk factors and generates the respective alerts.

Application of appropriate thresholds for each segment

The segmentation model offered by the module allows the application of more accurate values to customer groups by identifying more specific groups and using statistical methodology to gain a better understanding of their behavior.

Accurate alerts and reduction of false positives

Monitor Plus® uses an assembly of expert models and AI to optimize the high level of detection and decrease false positives, resulting in a lower volume of alerts and a reduction in the operational burden placed on the analysis team. In addition, the increased accuracy of alerts reduces customer friction and improves customer satisfaction.

Monitor Plus K-Prototype Segmentation™ analyzes numerical and categorical variables to maximize the information available to understand customer behavior and detect unusual transactions that potentially need to be reported to the regulator.

 

The system performs an exploratory analysis of the data to provide descriptive knowledge about the quality, distribution and completeness of the variables analyzed. This allows the system to identify the amount of null data, maximums, minimums, the means of numerical variables and the number of categories included in the categorical variables.

Analysis of numerical and categorical variables:

Maximization of available information

Exploratory analysis allows the model to remove outliers from its analysis to avoid errors in the model. This allows the system to correlate data to identify those with very similar behaviour and those that do not add value or information to the model, determining the characteristics and numbers of segments according to risk factors (persons, products, channels, jurisdictions) and the subsequent assignment of the risk level of each segment.

 

Every client is monitored periodically against their segment to analyze whether there were segment changes among their factors and the system generates alerts when the customer's behavior exceeds the risk thresholds established by the institution.