Customer Churn Heatmap
Objective: The standard churn prediction models in telecom rely on variables related to service usage and billing data. The drawback of these variables is they are lagging indicators. CRM interaction variables, on the other hand, are leading indicators of churn. The objective is to enhance the standard churn prediction model using CRM interaction variables. Approach: The affinity between CRM interaction variables (viz. repeat interactions, complaints on specific service categories etc) and the propensity to churn was mined using the priori algorithm. The Confidence of the association rule is used as the probability of churn; the Lift is used to prioritize the rule. The resulting model is made available to the end users (Retention Managers) on Tableau as an interactive Heatmap (please refer the attached snapshot)
เกี่ยวกับฉัน
I have 18 yoe in ML, Big Data, BI and related data science works. Focused on helping a few niche technology companies in bringing their (mostly disruptive) ideas to life. A design thinker, responsive and collaborative individual with strong background in data science. My clients engage me to architect and develop solutions that have Big Data and/or Machine Learning as differentiating components. - Predictive Analytics & ML - Big Data Backends - Streaming Data Pipelines - Good Old ETL & BI Machine Learning: Regression, Association (apriori), Classification (decision trees, random forest, logit), Clustering (k-means) Python, Scikit-learn Dataiku, RapidMiner, Azure ML Studio, SageMaker Data Visualisation: Power BI, Tableau, Qlik, Klipfolio, Kibana DW and ETL: Snowflake, BigQuery, Athena, AirFlow, SSIS Stream Processing: Spark, Flume, Kafka, Kafka Streams, Kinesis Elasticsearch (ELK) I believe in simple and future-proof design. Putting trust and satisfaction before money.
