PICK READY-TO-USE MODEL

Kickstart your process by taking advantage of a Ready-to-Use churn Model.

For getting the relevant list of ready to-use models, pick 'Keep' from the Customer Lifecycle Stage dropdown.

PICK READY-TO-USE MODEL

For building a model for churn prediction, choose 'ChurnStatus' from the 'ScoreName' dropdown.

MULTIPLE MODEL TYPES

And to be more precise of the customer type, choose 'PrePaid' Subscriber Type.

HERE'S YOUR MODEL, SIR!

In a jiffy, utilizing the deep subject matter expertise of our CVM experts, you have in front of you the list of features (or attributes) that have the maximum impact on the 'ChurnStatus’. Download this precious features catalog to get the necessary data!

Time to Explore your data!

GET YOUR DATA

Choose the source of your data : a local file, a connected database, or data in AI^x server

AUTO EXPLORE RESULTS

Automatically the most relevant feature pair plots are built to explore and understand the data better.

DATA SUMMARY CREATED

You can download and review summary statistics for all features for data quality assessment and to surface common data quality issues.

VIDEO TOUR

Manual Exploration

VISUALIZE AS YOUR LIKE

Moreover, let your nerdy mind play and manually perform the data analysis, a whole bunch of visualizations are available - Histogram, BarGraph, PieChart, StripPlot….all you can ask for!

VIDEO TOUR

Data Exploration

CREATE NEW MODEL

In MAI^x, you can start by creating a fresh new model or modifying an earlier built model.

NAME YOUR MODEL

Let's give a name for our new project to create a prediction model

GET YOUR DATA

Choose the source of your data : a local file, a connected database, or data in AI^x server

PREVIEW YOUR DATA

A preview of the your chosen data comprising of sample 50 rows or records

SELECT DATASCIENCE USECASE

Pick the machine learning approach, Supervised - for classification or regression problems

Unsupervised - for clustering, association or dimensionality reduction problems

Choose 'Supervised' for our example of churn prediction (which is a classification problem).

TRAIN YOUR MODEL

Multi-Select all the features (or column names) of your data that you believe impact the churn of a customer.

The selected features will be the input for training the model.

WHAT WOULD YOU LIKE TO PREDICT?

The target feature is the name of the column in the dataset that you would like to predict. Examples include:

- "ChurnStatus": has this customer churned out?

- "Is_Satisfied": Is the customer satisfied ?

- "Likes_Personalized_Offers": Does the customer likes to receive personalized offers?

VIDEO TOUR - APPLY CUSTOM SETTINGS

You can customize model building by setting advanced options. However, MAIx applies the best suited settings that provide comprehensive insights without additional configuration.

VIDEO TOUR

Model Result

EUREKA! THE MOST ACCURATE MODEL

In the end, AI^x ranks and suggests the top most accurate model and prepares it for deployment. If you want to deploy an alternate, it's easy to prepare a different model.

VIDEO TOUR

Model Result

FEATURES IMPORTANCE

See the red color bars on the left? Those are the feature importance scores. They indicate a feature's correlation with the target ("ChurnStatus") and serve as a measure of predictive power.

VIEW, APPLY OR DOWNLOAD

Manage all your built models: view or transfer model files, apply model output to production environment or even download the developed model code

DEPLOYING A MODEL

You may choose to deploy a model, modify an existing code base, schedule a chrontab to automate your task or or even use Rest APIs

APPLY DEPLOYMENT SETTINGS

You can customize model deployment by choosing the options. However, MAI^x applies the best suited settings that provide quick deployment without additional configuration.

VIDEO TOUR - SETUP CRONTAB, REST API

Configuring and modifying CronTabs and Managing REST API deployment comes easy and quick

PRODUCTION DASHBOARD

Dashboard presents a summary of the running production jobs, their active status and quick access their logs.

MONITOR MODEL PERFORMANCE

See the performance for all the running models or filter to see performance results for a time duration or select model.

ANALYSE AT DEEPER LEVEL

Compare the performance of the model on training vs testing dataset, and see more in-depth analysis

VIDEO TOUR - WANT MORE ANALYSIS , EH?

Don't miss out on the study of response rates by quantile, decile, score group, and the auto created confusion matrices.