<

• Site Navigation
  • Project Home
• Analyst's Notebooks
  • EDA Cleanse - .ipynb
  • EDA Analysis - .ipynb
  • Feature Engineering - .ipynb
• ML Models
  • Logistical Regression - .ipynb
  • Decision Tree - .ipynb
  • XGBoost - .ipynb
  • Random Forest - .ipynb
  • Model Comparisons - .ipynb
  • Model Demonstration - .ipynb
• Client Documentation
  • Client Proposal - Docs
  • PACE Plan
  • Analyst's Summary
  • Executive Summary Analysis
  • HR Summary Analysis
  • Team Summary Generator Notebook
  • Team Summary example
  • Project Close
• Project Conclusion Presentations
  • Executive Summary PRESENTATION
  • HR Summary PRESENTATION

Salifort Motors - Demonstration model

Document Information

Document Title	Salifort Motors - Demonstration Model - Best Choice Random Forest
Author	Rod Slater
Version	1.0
Created	01-11-2023
Modified	16-11-2023

Client Details

Client Name	Salifort Motors
Client Contact	Mr HR Team
Client Email	hr@salifortmotors.it
Client Project	HR Team Data Driven Solutions from Machine Learning Models

Document Overview

Demonstration of the ML models created for the Salifort Motors project.

• "will_the_leave" - Interactive function to send employee data to that will output a prediction from the three models.
• Summary by risk - a summary of the predictions from the three models
• Predictions file - that applies the three models to the HR data and outputs the data with prediction added for each employee row.
• Summary of the three models

Table of contents

• Salifort Motors - Demonstration model
  • Document Information
  • Client Details
  • Document Overview
• Initialise Notebook
  • Import Packages
  • Initialise Notebook Options
  • Load the dataset
  • Pickle Load models developed using the Focus Features dataset
• Predict cost of salary increase for employees predicted to leave by the ML Model
• Predictions Summary of the model
• will_they_leave function- Focus Features
  • Interactive Model Comparison Demonstration
• Model prediction Summaries
  • Load and Prepare data
  • Set Prediction Probability Thresholds for reporting
  • Summary - Random Forest Model Predictions
  • Summary XGBoost Predictions
  • Summary Decision Tree Predictions
  • Summarise model findings
• View Results in browser

Initialise Notebook

Since there are no substantial variances in predictions, accuracy or any other score for the two model datasets (AllFeat - All Features, and NoDept - Departments removed, we'll use the NoDept models since, there are less variables to type out!)

Import Packages

import pickle
import pandas as pd

# Add a little colour

class color:
   PURPLE = '\033[95m'
   CYAN = '\033[96m'
   DARKCYAN = '\033[36m'
   BLUE = '\033[94m'
   GREEN = '\033[92m'
   YELLOW = '\033[93m'
   RED = '\033[91m'
   BOLD = '\033[1m'
   UNDERLINE = '\033[4m'
   END = '\033[0m'

Initialise Notebook Options

load_path = "/home/hass/Documents/Learning/Salifort-Motors-Capstone-Project/00-data_cleaned/" # Source folder for cleaned data
#save_path = "/home/hass/Documents/Learning/Salifort-Motors-Capstone-Project/04-pickle-ML-models/" # destination for pickle saved models

# path to load/save pickled models
model_path = "/home/hass/Documents/Learning/Salifort-Motors-Capstone-Project/04-pickle-ML-models/" 

# path to save the predictions from  each model applied to the HR datafile
predictions_path = "/home/hass/Documents/Learning/Salifort-Motors-Capstone-Project/99-documentation-project/07-predictions/" 

Load the dataset

NoDept_df = pd.read_csv(load_path + "data_cleaned_NoOl_FE_NoDept.csv", index_col = False) 
NoDept_df.sort_index(axis = 1, inplace=True)

NoDept_df.columns

Index(['last_eval', 'left', 'number_project', 'overworked', 'promotion',
       'salary', 'satisfaction', 'tenure'],
      dtype='object')

Pickle Load models developed using the Focus Features dataset

# Load the trained  Random Forest Model
with open(model_path + '/hr_rf2-NOdept.pickle', 'rb') as model_file_rf:
    model_file_rf = pickle.load(model_file_rf)

# Load the trained Decision Tree model
with open(model_path + '/hr_dt2-NOdept.pickle', 'rb') as model_file_dt:
    model_file_dt = pickle.load(model_file_dt)

# Load the trained XGBoost model
with open(model_path + '/hr_xg2-NOdept.pickle', 'rb') as model_file_xg:
    model_file_xg = pickle.load(model_file_xg)

Predict cost of salary increase for employees predicted to leave by the ML Model

# Review costs of salary ranges for staff predicted to leave
def predict_salary_increase(modelname, title, salary_inc_perc): 

      '''    Generate summary table of costs of a salary increase for predicted employee will leave

      In: 
            modelname: str      : String : name of the model file e.g. model_file_xxx 
            title: str          : String :for the summary title, usually full description of model e.g. 'Decision Tree', 'XGBoost', 'Random Forest'
            salary_inc_perc     : Float  : percentage to increase salary by e.g. 0.50
      Out:  print of the summary
    '''

      # Set output number format
      pd.set_option('display.float_format', '{:.2f}'.format)

      df = AllFeat_df.copy()

      count_by_group = df.groupby(['salary', 'left']).size().reset_index(name='count')

      # Select employees stayed (left=0) and employees who left (left=1)
      features_current_empl = NoDept_df[NoDept_df['left'] == 0][['last_eval', 'number_project',
            'overworked', 'promotion', 'salary', 'satisfaction', 'tenure']]

      #print(features.head())

      # Predictions for CURRENT EMPLOYEES using model_file_{modelname}

      if modelname == "model_file_dt":
            probabilities_cur = model_file_dt.predict_proba(features_current_empl)
      elif modelname == "model_file_xg":
            probabilities_cur = model_file_xg.predict_proba(features_current_empl)
      elif modelname == "model_file_rf":
            probabilities_cur = model_file_rf.predict_proba(features_current_empl)

      leave_probabilities_cur = probabilities_cur[:, 1]
      #stay_probabilities_cur =  probabilities_cur[:, 0]
      #print(features.shape)
      #print(leave_probabilities.shape)

      # Add the probabilities to the new data DataFrame if needed
      features_current_empl['leave_probability'] = leave_probabilities_cur
      #features_current_empl['stay_probability'] = stay_probabilities_cur

      # Sort the df
      predictions_cur = features_current_empl.sort_values(by='leave_probability', ascending=False)

      # predictions_cur
      # Summarise predictions

      high_risk_count_cur = len(predictions_cur[predictions_cur['leave_probability'] > high_risk_threshold])
      high_risk_perc_cur = (high_risk_count_cur / len(predictions_cur)) * 100
      medium_risk_count_cur = len(predictions_cur[predictions_cur['leave_probability'] > medium_risk_threshold])
      medium_risk_perc_cur = (medium_risk_count_cur / len(predictions_cur)) * 100
      low_risk_count_cur = len(predictions_cur[predictions_cur['leave_probability'] > predict_true_threshold])
      predict_risk_perc_cur = (low_risk_count_cur / len(predictions_cur)) * 100

      # Display the result
      # print(count_by_group)

      high_left   = df[(df['salary'] == 'high') & (df['left'] == 0)]
      medium_left = df[(df['salary'] == 'medium') & (df['left'] == 0)]
      low_left    = df[(df['salary'] == 'low') & (df['left'] == 0)]


      # Initialise values

      high_salary   = 150000
      medium_salary = 100000
      low_salary    =  50000

      # salary_increase_percentage = 0.25 

      # Count the rows in the filtered DataFrame multiply by salary thresholds

      # High salary and salary increase 
      high_cost = high_risk_count_cur * high_salary * salary_inc_perc

      # Assumes medium salary @ £100K and cost to replace is 30% of salary
      medium_cost = medium_risk_count_cur * medium_salary * salary_inc_perc

      # Assumes low salary @ £50K and cost to replace is 30% of salary
      low_cost = low_risk_count_cur * low_salary * salary_inc_perc

      # Count of employees
      # total_count = low_left.shape[0]+medium_left.shape[0]+high_left.shape[0]
      total_count = high_risk_count_cur + medium_risk_count_cur + low_risk_count_cur


      print()
      # Display the selected rows

      print(color.BOLD + f"{title} - Cost of salary increase of {salary_inc_perc:.0%} of total salary)" + color.END)
      print('\u2500' * 70)
      print("                                  |   # of Staff |  Cost Per Head |   Est £ Cost")
      print("Estimate High Salary Left Cost    | {:>12,.0f} | {:>14,.0f} | {:>12,.0f} ".
            format(high_risk_count_cur, high_cost/high_risk_count_cur, high_cost))
      print("Estimate Medium Salary Left Cost  | {:>12,.0f} | {:>14,.0f} | {:>12,.0f}"
            .format(medium_risk_count_cur, medium_cost/ medium_risk_count_cur, medium_cost))
      print("Estimate Low Salary Left Cost     | {:>12,.0f} | {:>14,.0f} | {:>12,.0f}"
            .format(low_risk_count_cur, low_cost/ low_risk_count_cur, low_cost))
      print("                                  | ============ |         ====== |    =========")
      print("Estimate TOTAL Salary Left Cost   | {:>12,.0f} |                | {:>12,.0f} "
            .format(total_count, low_cost+medium_cost+high_cost))

#predict_salary_increase('model_file_xg', "Test", 0.50)

Predictions Summary of the model

def model_predictions(modelname, title):
       '''    Generate summary table for predictions if an employee will leave and apply predictions to employees left to compare

    In: 
        modelname: str      : model_file_xxx name of the model file string. 
        title: str          : String for the summary title, usually full description of model e.g. 'Decision Tree', 'XGBoost', 'Random Forest'
        

    Out: print of the summary
    '''
       for i in range(2):

              # Select employees stayed (left=0) and employees who left (left=1)
              features_current_empl = NoDept_df[NoDept_df['left'] == i][['last_eval', 'number_project',
                     'overworked', 'promotion', 'salary', 'satisfaction', 'tenure']]

              #print(features.head())

              # Predictions for CURRENT EMPLOYEES using model_file_xg for XGBoost
              if modelname == "model_file_dt":
                     probabilities_cur = model_file_dt.predict_proba(features_current_empl)
              elif modelname =="model_file_rf":
                     probabilities_cur = model_file_rf.predict_proba(features_current_empl)
              elif modelname == "model_file_xg":
                     probabilities_cur = model_file_xg.predict_proba(features_current_empl)

              leave_probabilities_cur = probabilities_cur[:, 1]
              stay_probabilities_cur =  probabilities_cur[:, 0]
              #print(features.shape)
              #print(leave_probabilities.shape)

              # Add the probabilities to the new data DataFrame if needed
              features_current_empl['leave_probability'] = leave_probabilities_cur
              features_current_empl['stay_probability'] = stay_probabilities_cur

              # Sort the df
              predictions_cur = features_current_empl.sort_values(by='leave_probability', ascending=False)

              # predictions_cur
              # Summarise predictions

              high_risk_count_cur = len(predictions_cur[predictions_cur['leave_probability'] > high_risk_threshold])
              high_risk_perc_cur = (high_risk_count_cur / len(predictions_cur)) * 100
              medium_risk_count_cur = len(predictions_cur[predictions_cur['leave_probability'] > medium_risk_threshold])
              medium_risk_perc_cur = (medium_risk_count_cur / len(predictions_cur)) * 100
              low_risk_count_cur = len(predictions_cur[predictions_cur['leave_probability'] > predict_true_threshold])
              predict_risk_perc_cur = (low_risk_count_cur / len(predictions_cur)) * 100

              if i == 0:
                     print("\n"+ color.BOLD + f"{title} Prediction Summary for CURRENT Employees\n" + color.END)
                     print('\u2500' * 55) 
              else:
                     print("\n\n"+ color.BOLD + f"{title} Prediction Summary for LEFT Employees (how good is the model?)\n"+ color.END)
                     print('\u2500' * 55) 

       # Print the result
              print(f"\n" + color.BOLD + f"probabilities > high_risk_threshold {high_risk_threshold:.0%}" + color.END)
              print(f'Count of employees with leave probability above {high_risk_threshold:.0%}        : {high_risk_count_cur:.0f}')
              print(f'Percentage of employees with leave probability above {high_risk_threshold:.0%}   : {high_risk_perc_cur:.2f}%')

              print(f"\n" + color.BOLD +  f"probabilities > medium_risk_threshold {medium_risk_threshold:.0%}" + color.END)
              print(f'Count of employees with leave probability above {medium_risk_threshold:.0%}        : {medium_risk_count_cur:.0f}')
              print(f'Percentage of employees with leave probability above {medium_risk_threshold:.0%}   : {medium_risk_perc_cur:.2f}%')

              print(f"\n" + color.BOLD + f"probabilities > low_risk_threshold {predict_true_threshold:.0%}" + color.END)
              print(f'Count of employees with leave probability above {predict_true_threshold:.0%}        : {low_risk_count_cur:.0f}')
              print(f'Percentage of employees with leave probability above {predict_true_threshold:.0%}   : {predict_risk_perc_cur:.2f}%')
              
              if i == 0:
                     pass
              else:
                     print(f"\n" + color.BOLD + color.UNDERLINE + "\n% of employees left that were predicted\n" + color.END)
                     #print('\u2500' * 55) 
                     print(f'XGBoost Predicted to leave > ({predict_true_threshold:.0%} )/ % of employees who left         : {low_risk_count_cur/1991*100:.2f} %')


       predictions_cur.to_html(predictions_path + "Predictions_XGBoost.html")

will_they_leave function- Focus Features

Function to demonstrate the different predictions from the three models Random Forest, Decision Tree and XGBoost developed using the FOCUS features dataset. This dataset EXCLUDES the dummy encoded dept data.

def will_they_leave_NoDept(satisfaction:float, last_eval:float, num_proj:int, tenure:int, overworked:bin, promotion:bin, salary:int):
    '''
    Generate predictions if an employee will leave, with a probability

    In: 
        last_eval:          Last evaluation score 0-1 float
        num_proj:           Number of projects them employee has worked on 1-7 int
        tenure:             Number of years employed 2-10 int 
        overworked          If an employee works more than 175 hours a month bin

    Out: String statement of prediction and % probability
    '''

    

    input_data2 = {
        'last_eval'         : last_eval,

        'number_project'    : num_proj, 
        'overworked'        : overworked,
        'promotion'         : promotion,  # 1 for yes, 0 for no
        'salary'            : salary,     # 0 =low, 1 = medium, 2 = high
        'satisfaction'      : satisfaction,
        'tenure'            : tenure     
        #'dept_IT'           : 0, 
        #'dept_RandD'        : 0,
        #'dept_accounting'   : 0, 
        #'dept_hr'           : 1, 
        #'dept_management'   : 0, 
        #'dept_marketing'    : 0,
        #'dept_product_mng'  : 0, 
        #'dept_sales'        : 0, 
        #'dept_support'      : 0, 
        #'dept_technical'    : 0,
        
        
    }

    # Convert the input data to a DataFrame
    input_df2 = pd.DataFrame(input_data2, index=[0])

    # Make a prediction
    predicted_class = model_file_rf.predict(input_df2)
    probability = model_file_rf.predict_proba(input_df2)[:, 1]

    # Output the prediction and probability
    print()
    print(color.BOLD + "Random Forest Model 2 - No departments" + color.END)
    print('\u2500' * 40) 
    if predicted_class[0] == 1:
        print("The employee is **LIKELY** to leave with a probability of {:.2f}%".format(probability[0] * 100))
    else:
        print("The employee is not likely to leave with a probability of {:.2f}%".format((1 - probability[0]) * 100))

    predicted_class = model_file_dt.predict(input_df2)
    probability = model_file_dt.predict_proba(input_df2)[:, 1]

    print()
    print(color.BOLD + "Decision Tree Model 2 - No departments" + color.END)
    print('\u2500' * 40) 
    if predicted_class[0] == 1:
        print("The employee is **LIKELY** to leave with a probability of {:.2f}%".format(probability[0] * 100))
    else:
        print("The employee is not likely to leave with a probability of {:.2f}%".format((1 - probability[0]) * 100))

    predicted_class = model_file_xg.predict(input_df2)
    probability = model_file_xg.predict_proba(input_df2)[:, 1]

    print()
    print(color.BOLD + "XGBoost Model 2 - No departments" + color.END)
    print('\u2500' * 40) 
    if predicted_class[0] == 1:
        print("The employee is **LIKELY** to leave with a probability of {:.2f}%".format(probability[0] * 100))
    else:
        print("The employee is not likely to leave with a probability of {:.2f}%".format((1 - probability[0]) * 100))

Interactive Model Comparison Demonstration

# Amend the variables below to create an employee record example to test if a employee is predicted to leave
# Note Key Feature : tenure, promotion, num_projects, last_eval.

number_of_projects      = 1         # Discrete range [0-7]
salary                  = 1         # 0 = low, 1 = medium, 2 = high
satisfaction            = 0.36       # Continuous [0-1] demo example 0.36 predicts to leave, 0.7 predicts to stay
last_evaluation_score   = 0.46       # Continuous range [0-1]
employee_tenure         = 3         # Discrete range [2-10]
promotion_l5y           = 0         # Binary range [0,1] 
employee_overworked     = 0         # Binary [0,1] 
avg_mnth_hrs            = 300       # Discrete range [96 - 310] 175 hrs = overworked

will_they_leave_NoDept(satisfaction, last_evaluation_score, number_of_projects, employee_tenure, employee_overworked, promotion_l5y, salary)

Random Forest Model 2 - No departments
────────────────────────────────────────
The employee is **LIKELY** to leave with a probability of 97.09%

Decision Tree Model 2 - No departments
────────────────────────────────────────
The employee is **LIKELY** to leave with a probability of 95.89%

XGBoost Model 2 - No departments
────────────────────────────────────────
The employee is **LIKELY** to leave with a probability of 96.11%

Model prediction Summaries

• Add some highlighting
• Load the employee data file
• applies the Decision Tree, RandomForest and XGBoost models to the data
• and adds a prediction probability column that reflects the model's prediction probability that an employee is at risk from leaving.
• This is then iterated for all current employees (left = 0) and all employees who have left (left = 1) to provide a comparison
• The left = 1 prediction is a count of how many predictions the model made,

Load and Prepare data

# Load cleaned dataset into a dataframe
df1 = pd.read_csv(load_path + "data_cleaned_NoOl_FE_AllFeat.csv")

AllFeat_df = df1.copy()
NoDept_df = pd.read_csv(load_path + "data_cleaned_NoOl_FE_NoDept.csv")

AllFeat_df.sort_index(axis=1, inplace=True)
NoDept_df.sort_index(axis=1, inplace=True)

#print(AllFeat_df.columns)
#print(NoDept_df.columns)

AllFeat_LEFT_df = AllFeat_df[AllFeat_df['left'] == 1].drop(columns=['left'])

NoDept_LEFT_df = NoDept_df[NoDept_df['left'] == 1].drop(columns=['left'])

#print(AllFeat_LEFT_df.shape[0])
#print(NoDept_LEFT_df.shape[0])

Set Prediction Probability Thresholds for reporting

# Set risk thresholds for calcs

high_risk_threshold = 0.9
medium_risk_threshold = 0.7
predict_true_threshold = 0.5

Summary - Random Forest Model Predictions

model_predictions('model_file_rf', 'Random Forest')
predict_salary_increase('model_file_rf','Random Forest', 0.50) # modelname and model title

Random Forest Prediction Summary for CURRENT Employees

───────────────────────────────────────────────────────

probabilities > high_risk_threshold 90%
Count of employees with leave probability above 90%        : 26
Percentage of employees with leave probability above 90%   : 0.28%

probabilities > medium_risk_threshold 70%
Count of employees with leave probability above 70%        : 48
Percentage of employees with leave probability above 70%   : 0.52%

probabilities > low_risk_threshold 50%
Count of employees with leave probability above 50%        : 59
Percentage of employees with leave probability above 50%   : 0.64%


Random Forest Prediction Summary for LEFT Employees (how good is the model?)

───────────────────────────────────────────────────────

probabilities > high_risk_threshold 90%
Count of employees with leave probability above 90%        : 1704
Percentage of employees with leave probability above 90%   : 90.54%

probabilities > medium_risk_threshold 70%
Count of employees with leave probability above 70%        : 1715
Percentage of employees with leave probability above 70%   : 91.13%

probabilities > low_risk_threshold 50%
Count of employees with leave probability above 50%        : 1721
Percentage of employees with leave probability above 50%   : 91.45%


% of employees left that were predicted

XGBoost Predicted to leave > (50% )/ % of employees who left         : 86.44 %


Random Forest Model Predictions of salary cost increases

Cost of salary increase of 50% of total salary)
──────────────────────────────────────────────────────────────────────
                                  |   # of Staff |  Cost Per Head |   Est £ Cost
Estimate High Salary Left Cost    |           26 |         75,000 |    1,950,000 
Estimate Medium Salary Left Cost  |           48 |         50,000 |    2,400,000
Estimate Low Salary Left Cost     |           59 |         25,000 |    1,475,000
                                  | ============ |         ====== |    =========
Estimate TOTAL Salary Left Cost   |          133 |                |    5,825,000 

Summary XGBoost Predictions

model_predictions('model_file_xg', 'XGBoost')
predict_salary_increase('model_file_xg','XGBoost', 0.50) # modelname and model title

XGBoost Prediction Summary for CURRENT Employees

───────────────────────────────────────────────────────

probabilities > high_risk_threshold 90%
Count of employees with leave probability above 90%        : 12
Percentage of employees with leave probability above 90%   : 0.13%

probabilities > medium_risk_threshold 70%
Count of employees with leave probability above 70%        : 33
Percentage of employees with leave probability above 70%   : 0.36%

probabilities > low_risk_threshold 50%
Count of employees with leave probability above 50%        : 67
Percentage of employees with leave probability above 50%   : 0.72%


XGBoost Prediction Summary for LEFT Employees (how good is the model?)

───────────────────────────────────────────────────────

probabilities > high_risk_threshold 90%
Count of employees with leave probability above 90%        : 1360
Percentage of employees with leave probability above 90%   : 72.26%

probabilities > medium_risk_threshold 70%
Count of employees with leave probability above 70%        : 1712
Percentage of employees with leave probability above 70%   : 90.97%

probabilities > low_risk_threshold 50%
Count of employees with leave probability above 50%        : 1731
Percentage of employees with leave probability above 50%   : 91.98%


% of employees left that were predicted

XGBoost Predicted to leave > (50% )/ % of employees who left         : 86.94 %


XGBoost Model Predictions of salary cost increases

Cost of salary increase of 50% of total salary)
──────────────────────────────────────────────────────────────────────
                                  |   # of Staff |  Cost Per Head |   Est £ Cost
Estimate High Salary Left Cost    |           12 |         75,000 |      900,000 
Estimate Medium Salary Left Cost  |           33 |         50,000 |    1,650,000
Estimate Low Salary Left Cost     |           67 |         25,000 |    1,675,000
                                  | ============ |         ====== |    =========
Estimate TOTAL Salary Left Cost   |          112 |                |    4,225,000 

Summary Decision Tree Predictions

model_predictions('model_file_dt', 'Decision Tree')
predict_salary_increase('model_file_dt','Decision Tree', 0.50) # modelname and model title

Decision Tree Prediction Summary for CURRENT Employees

───────────────────────────────────────────────────────

probabilities > high_risk_threshold 90%
Count of employees with leave probability above 90%        : 40
Percentage of employees with leave probability above 90%   : 0.43%

probabilities > medium_risk_threshold 70%
Count of employees with leave probability above 70%        : 87
Percentage of employees with leave probability above 70%   : 0.94%

probabilities > low_risk_threshold 50%
Count of employees with leave probability above 50%        : 87
Percentage of employees with leave probability above 50%   : 0.94%


Decision Tree Prediction Summary for LEFT Employees (how good is the model?)

───────────────────────────────────────────────────────

probabilities > high_risk_threshold 90%
Count of employees with leave probability above 90%        : 1342
Percentage of employees with leave probability above 90%   : 71.31%

probabilities > medium_risk_threshold 70%
Count of employees with leave probability above 70%        : 1730
Percentage of employees with leave probability above 70%   : 91.92%

probabilities > low_risk_threshold 50%
Count of employees with leave probability above 50%        : 1730
Percentage of employees with leave probability above 50%   : 91.92%


% of employees left that were predicted

XGBoost Predicted to leave > (50% )/ % of employees who left         : 86.89 %


Decision Tree Model Predictions of salary cost increases

Cost of salary increase of 50% of total salary)
──────────────────────────────────────────────────────────────────────
                                  |   # of Staff |  Cost Per Head |   Est £ Cost
Estimate High Salary Left Cost    |           40 |         75,000 |    3,000,000 
Estimate Medium Salary Left Cost  |           87 |         50,000 |    4,350,000
Estimate Low Salary Left Cost     |           87 |         25,000 |    2,175,000
                                  | ============ |         ====== |    =========
Estimate TOTAL Salary Left Cost   |          214 |                |    9,525,000 

Summarise model findings

Count the number of employees that are at risk of leaving with probabilities set at:

• High Risk > 70%

• Medium Risk > 50% < 70%

• Low Risk < 50%

• XGBoost Predicted to leave > (50% )/ % of employees who left : 86.94 %

• Decision Tree Predicted to leave > (50% )/ % of employees who left : 86.89 %

• Random Forest Predicted to leave > (50% )/ % of employees who left : 86.44 %

View Results in browser

View XGBoost Predictions.html

View Decision Tree Predictions.html

View Random Forest Predictions.html