CRISP-DM Phase: Deployment of Solution in the GE Employee Attrition

CRISP-DM
Author
Affiliation
Course
Instructor
Due Date
Deployment
Deployment refers to the process of taking a trained Machine Learning (ML) model and making its predictions available to users or other systems. When a machine learning model's insights are frequently made available to the people for whom it was designed, it can begin to offer value to an organization. Because data scientists and machine learning engineers lack the technical expertise required in deployment, they must work alongside software engineers. First, it is important to understand how end-users should interact with the model's predictions before deciding how to deploy it.
The information is saved in a bucket on Microsoft Azure in two different file formats. The first component is a CSV file that describes the features of the data, and the second component is the dataset itself. It is possible that it will be more convenient for you to develop your machine learning system in the cloud, given that the data is already stored there. Nearly 1200 individuals are employed across the globe by GE. Important decisions about human resources are typically made, and payments to employees are dispersed at the end of each calendar month. At the end of each month, a compilation of the data is performed, and it is then saved in a cloud bucket. As a result, employment data isn't as extensive.
The fact that the data is hosted in a Microsoft Azure bucket makes it simple for models developed on the Microsoft Azure cloud platform to retrieve and utilize the data. GE will request data for inference through their proprietary application programming interface (API), while data for prediction will be obtained through a representational state transfer application programming interface (REST API) (Siam et al., 2021). A prediction request can be stored in a centralized database. The human resource staff would be the immediate use of the ML model's prediction score. A business intelligence tool would then be used for reporting to the end-users.
We will monitor the model's state and actively track it to look for any indications of depreciation, bias, or even data distortion and drift. This will ensure that such problems are immediately fixed before the end-user notices.
The cost of deploying the ML model is affordable. Cloud computing services provide price their services competitively. As a result, most of the resources are affordable. Microsoft and Google, for instance, charge users based on the storage spaces and the resources one needs. Organizations can choose fixed packages or subscribe to packages that vary depending on the resources consumed. Resources here include; computing power, storage, and other re...