System and method for predicting and reducing subscriber churn

A method for predicting and reducing subscriber churn on a computer network, the method comprising:

• retrieving traffic flow data for a subscriber of the computer network;
• determining at least one subscriber metric from the traffic flow data, wherein determining the at least one subscriber metric comprises:
• determining whether the at least one subscriber metric includes greater than a predetermined number of data points, and, if so:
• grouping the data points based on the time the metric was retrieved;
• taking a mean of the grouped data points;
• and
• using the mean of the grouped data points as the at least one subscriber metric;
• analyzing the at least one subscriber metric with a model for predicting subscriber churn;
• predicting whether the subscriber will churn during a predetermined churn period;
• and
• if the subscriber is predicted to churn, providing a traffic action on the traffic flow for the subscriber;
• otherwise allowing the subscriber's traffic flow to continue without action.

A system for predicting and reducing subscriber churn on a computer network, the system comprising:

• at least one processor connected to at least one memory storing instructions executable by the at least one processor to implement a plurality of modules comprising:
• a data collection module configured to:
• retrieve traffic flow data per subscriber for a plurality of subscribers in the computer network;
• and
• determine at least one systemic feature associated with the plurality of subscribers;
• a feature extraction module configured to determine at least one metric per subscriber form the traffic flow data, wherein determining the at least one subscriber metric comprises:
• determining whether the at least one subscriber metric includes greater than a predetermined number of data points, and, if so:
• grouping the data points based on the time the metric was retrieved;
• taking a mean of the grouped data points;
• and
• using the mean of the grouped data points as the at least one subscriber metric;
• a machine learning module configured to:
• analyze at least one metric and at least one feature for the predetermined time period;
• create a model for predicting and reducing subscriber churn based on the analysis;
• and
• predict per subscriber, whether the subscriber is going to churn within a churn period in the future based on the analysis;
• an evaluation module configured to validate the prediction by determining whether the subscriber actually churned during the churn period;
• and
• a reporting module configured to perform a traffic action based on the prediction.