Skip to content

Predict behavior to retain customers. You can analyze all relevant customer data and develop focused customer retention programs

Notifications You must be signed in to change notification settings

GaneshJainarain/TelcoCustomerChurn_AutoAI_ML

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

22 Commits
 
 
 
 
 
 

Repository files navigation

Telco Customer Churn

Context

Predict behavior to retain customers. You can analyze all relevant customer data and develop focused customer retention programs.

Content

Each row represents a customer, each column contains customer’s attributes described on the column Meta-data.

The data set includes information about:

  • Customers who left within the last month – the column is called Churn
  • Services that each customer has signed up for – phone, multiple lines, internet, online security, online backup, device protection, tech support, and streaming TV and movies
  • Customer account information – how long they’ve been a customer, contract, payment method, paperless billing, monthly charges, and total charges
  • Demographic info about customers – gender, age range, and if they have partners and dependents

Inspiration

To explore this type of models and learn more about the subject.

Downloading our dataset

Dataset

IBM Watson Studios

Watson Studio

Create a new project

Watson Studio

Creating a new asset AutoAI

Auto AI

Creating a new asset AutoAI II

Auto AI II

Predicting on Churn with Binary Classification

Predicting on Churn

As you can see as soon as we uploaded our dataset and selected what column we wanted to predict on, the experiment automatically gave us the best prediction type to use on the data, in this case Binary Classification and is optimized for accuracy and runtime

Progress Map of various Pipelines

Progress Map

Here we can see various amounts of pipelines, each one using a different algorithm, at the bottom we see a comparison of the pipelines and we can see that they choose the most optimal one for our task

Pipeline Comparison

Pipeline Comparison

Here we can see that Pipeline 4 has been chosen to be the best measured on accuracy, lets take a look inside the pipeline

Pipeline Analysis I

Pipeline Analysis I

Here we see the basic information of our Pipeline including the algorithm is used: Random Forest Classifier, the number of features: 34, and number of evaluation instances: 705.

Pipeline Analysis II

Pipeline Analysis II

Here we see the Feature Summary, each feature with its respective feature importance percentage.

Evaluation Metrics

Evaluation Metrics

Here we can see the Evaluation Metrics such as ROC Curve, Cross Validation Score, Precision, F1 Score, etc...

Confusion Matrix

Confusion Matrix

Here we see the Confusion Matrix

Precision Recall

Precision Recall

Saving our Model for Deployment

Model for deployment

Deployment of Model

Deployment of Model

Deployment of Model

Deployment of Model

import requests

# NOTE: you must manually set API_KEY below using information retrieved from your IBM Cloud account.
API_KEY = "<your API key>"
token_response = requests.post('https://iam.cloud.ibm.com/identity/token', data={"apikey":
 API_KEY, "grant_type": 'urn:ibm:params:oauth:grant-type:apikey'})
mltoken = token_response.json()["access_token"]

header = {'Content-Type': 'application/json', 'Authorization': 'Bearer ' + mltoken}
# NOTE: manually define and pass the array(s) of values to be scored in the next line
payload_scoring = {"input_data": [{"fields": [array_of_input_fields], "values": [array_of_values_to_be_scored, another_array_of_values_to_be_scored]}]}

response_scoring = requests.post('https://us-south.ml.cloud.ibm.com/ml/v4/deployments/41ccb00d-42ea-4b4a-ab21-5721eaeb9eb3/predictions?version=2023-02-28', json=payload_scoring,
 headers={'Authorization': 'Bearer ' + mltoken})
print("Scoring response")

About

Predict behavior to retain customers. You can analyze all relevant customer data and develop focused customer retention programs

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published