The client is a large MNC with 9 broad verticals across the organisation. The client faces a challenge in identifying the right people for promotion and preparing them in time. Under the current process, the final promotions are only announced after a first round of training and evaluation; and this leads to delay in employees' transition to their new roles. Hence, the company needs help in identifying the best candidates at an earlier juncture, in order to expedite the entire promotion cycle.

The company has provided the following employee datasets:

• Emp_ID: Unique ID for each employee

• Department: The department in which the employee works

• No of Region: Region of employment (unordered)

• Level of Education: Education Level

• Gender: Gender of employee

• Recruitment_channel: Channel through which employee was recruited

• No_of_trainings_completed: Number of other training sessions completed in previous year on soft skills, technical skills, etc.; but after which promotion did not occur.

• Age: Age of employee

• Performance score: Employee rating for the previous year

• Length_of_service: Length of service (in years)

• High KPIS: If percentage of KPIs > 80%, then 1; else 0.

• Awards won: If awards won during previous year, then 1; else 0.

• Average score evaluation: Average score in current training evaluations

• Is_promoted: Employee was actually promoted (this is our target variable). 1 if promoted; else 0.

Data cleaning and Data Preprocessing

• The "level of education" and "performance score" columns contained null values, which were addressed.

• Column labels were renamed for ease of understanding.

• Exploratory analysis was performed on the data, creating several charts to visualize the important trends. To see the Exploratory Data Analysis, click here

• Extract and create new features, also the reason behind them

Building the model

• Label Encoded all categorical features

• Create the target vector

• Split the data into training and testing with optional portion

• Apply "Standart ScaLer" from the sklearn preprocessing library to scale X_train dataset

• Explored the model performance of different ML classification algorithms. ( Logistic Regression, Random Forest Classifier , Gaussian Naive Bayes)

• Analysis of various metrics to determine the best algorithm, The metric used to decide the best algorithm was accuracy.

• The most promising algorithm (RFC) was chosen, and adjusted as necessary.

• The characteristics that have the greatest impact on the prediction were calculated.