Hybrid Machine Learning Models for Predictive Maintenance in Cloud-Based Infrastructure for SaaS Applications

Saransh Arora

Abstract:

It is crucial for SaaS providers to have predictive
maintenance as one of the most cost-effective ways for them to
successfully maintain service consistency and high satisfaction
among their customers. The present research article renders an
original technique to forecast the drop-offs of the clients in
Customer Relationship Management (CRM) systems through
the infrastructure of SaaS in the cloud. Working with a big
churn dataset obtained from Kaggle dataset created via a
telecom provider and then carefully investigating the feature
engineering, preprocessing, and data collecting. Next, some
machine learning methods are deployed with SVM + Naïve
Bayes model, KNN, DT, RF, and ANN. The model of Hybrid
SVM + Bayes performed better compared to the individual
models being based on the study results, the accuracy was
95.67%. It is revealed that the hybrid ML model shows a
significantly higher level of precision, accuracy, recall, and F1-
score than individual models do when compared through
thorough and methodological model training and evaluation.
The outcome emphasizes how effectual hybrid machine learning
algorithms are for SaaS arrangements to accomplish better
retention tactics in the dynamic world. This framework provides
a basis for future projects involving prediction maintenance that
are cloud-based SaaS systems as well as beneficial intelligence to
enterprises.