13
CHURN PREVENTION WITH ADVANCED ANALYTICS Sergio Sánchez [email protected] linkedin.com/in/sergiosanchezjorge

Churn prevention with advanced analytics

Embed Size (px)

Citation preview

CHURN PREVENTION WITH ADVANCED ANALYTICS

Sergio Sánchez

[email protected]

linkedin.com/in/sergiosanchezjorge

Analyze hundreds of anonymized clients from Santander Bank to predict if a customer is satisfied or dissatisfied with their banking experience.

OBJECTIVE

DATA SET

The bank database used here was taken from customers of Santander Bank. It contains information about 76019 customers of a bank with 370 attributes.

This adds up to a total of 28.127.030 data.

Neural Designer is the most advanced software for advanced analytics.

NEURAL DESIGNER

DATA DISTRIBUTION

We conducted a general analysis of data.

The churn rate here is 4.0%.

Evaluate which variables can influence churn.

VARIABLES IMPORTANCE

The máximum correlation (0,15046) is yield between the variable num_var30 and the target variable.

Configure the neural network architecture. This model takes the attributes of clients to predict their likelihood of being dissatisfied customers.

NEURAL NETWORK

LOSS INDEX

If the weighted squared error has a value of unity then the neural network ispredicting the data 'in the mean', while a value of zero means perfect predictionof the data.

The error term is the weighted squared error. It weights the squared error of negatives and positives values.

TESTS

The next step is to evaluate the performance of the trained neural network.

Confusion matrix

Binary classification

ROC curve

RESULTS

We apply the predictive model to 75819 potential customers.

The probability that client 1 is dissatisfied is 27%: the bank will not contact this customer.

The probability that client 4 is dissatisfied is 61%: the bank will contact this customer.

RESULTS

The predictive model can be easily integrated into the bank's computer system.

MODEL DEPLOYMENT

artelnics.comArtificial Intelligence Techniques, SL

Carretera de Madrid 1337900 Santa Marta de Tormes

Salamanca (Spain)Telephone: +34 923 133 612 Ext.13

E-mail: [email protected]

Supported by