PACE PLAN
Document Information
| Document Title | Salifort Motors - PACE Plan - for internal use only |
| 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 |
PACE - (P)LAN
PACE document (this document)
- Prepare a project proposal to define the project objective
- Prepare a PACE document outlining the project steps
Project Objective
Salifort Motors' Leadership and HR teams are concerned about the seemingly high number of staff leaving the company.
We have been employed to analyse their employee data in order to develop a ML model to predict the probability of an employee leaving.
Timeline
Project Deliverables
Agreed project deliverables:
Produce an Executive Summary of the data and project detailing key findings and recommendations
Executive Summary Presentation
Executive Summary Detailed Analysis
Produce an in depth analysts' report of findings and recommendations
Demonstration model :
- Interactive model for what if modelling
- Apply the model to current employee data to Identify employees “at risk” of leaving
Follow Up:
Support HR team & Team Leader with analysis to support workshops to review the data for their teams.
Team Summary Analysis Generator Notebook
Team Summary Generator Example output
Data Exploration and Cleaning
- Resources Required
- Stakeholder input
- Stakeholder Analysis
- Project Timeline (Gantt Chart)
- Dataset (HR_capstone_dataset.csv)
- Python Jupyter Notebooks
Python Libraries
| Operations | Python Libraries |
|---|---|
| Data Import & Manipulation |
Pandas Numpty |
| Data Modelling |
sklearn.linear_model - Logistic Regression sklearn.tree - Decision Tree sklearn.ensemble Random Forest XGBoost - |
| Modelling support and Metrics |
sklearn.model_selection GridSearch sklearn.metrics - model scoring sklearn.plot_tree - Decision Tree Visualisations |
| Visualisations |
matplotlib seaborn |
| ML Model Save/Load | pickle |
- Review the supplied data and apply cleaning operations to prepare it
for further analysis and modelling
- Deduplication
- Remove Missing or incomplete data
- Rename column headings for consistency
- Review Outliers and consider removing for relevant ML models
- Dataset Preparation:
- Because this is a capstone project and not real life, I will prepare two datasets to evaluate performance of these across the four models we'll evaluate for this project, because I'm curious how the two different datasets will perform.
- Prepare a dataset with minimal Feature Engineering and a dataset with comprehensive Feature Engineering
Analyse
Dataset overview
| Variable | Description |
| satisfaction_level | Employee-reported job satisfaction level [0–1] |
| last_evaluation | Score of employee's last performance review [0–1] |
| number_project | Number of projects employee contributes to |
| average_monthly_hours | Average number of hours employee worked per month |
| time_spend_company | How long the employee has been with the company (years) |
| Work_accident | Whether or not the employee experienced an accident while at work |
| left | Whether or not the employee left the company |
| promotion_last_5years | Whether or not the employee was promoted in the last 5 years |
| Department | The employee's department |
| salary | The employee's salary (low, medium, high) |
Exploratory Data Analysis
- Pairplot to identify key correlations
- Correlation Plot & Heatmap to identify key correlations and mean average scores
- Histogram plot for variable distribution
EDA Data Cleaning & Feature Engineering
- De-duplicate
- Remove missing values
- Check / correct any categorical value misspellings
- Correct any misspellings of column names, convert case of columns to snake case
- Removal of outliers if required by a specific ML Model
- As required feature engineering of categorical features to relevant encoding (dummies, ordinal), binarisation of continuous variables with threshold.
EDA
Analysis and visualisations of key data features Vs
left
- Variable comparison with
left:- Visualisation of the feature (histogram, heatmap, box plot, violin plot)
- Detailed Analysis of variable
- Further ad hoc visualisations and analysis will be produced driven by the
- variable comparison observations
Deliverable: EDA Analysis Summary
Client Deliverables
- Write up documenting initial findings from initial EDA
- Write up suggestions to HR team and Team Managers
- ML model performance Analysis comparisons
- ML Model Demonstration and prediction demonstration notebook
Project Deliverables-11
- Internal Analysis and RFC to analyst team members
Construct
Build, Train/Test
Build train and test various machine learning models:
- Logistic Regression
- Decision Tree
- XGBoost
- Random Forest
Disclaimer - in reality I wouldn't carry out the sort of overkill development you see here, but I want to compare model performance because I'm curious to see how they perform with feature engineering vs without any Feature Engineering
cleaned data
data_cleaned_NoOl_FE_AllFeat - Cleaned, No Outliers,
Feature Engineered, all fields included.- (AllFeat)
data_cleaned_NoOl_FE_NoDept - Cleaned, No Outliers, Feature
Engineered, departments removed (NoDept)
data_cleaned_NoOl_NoFE_AllFeat - Cleaned, No Outliers, NOT
Feature Engineered, all fields included.- (AllFeat)
data_cleaned_Ol_NoFE_AllFeat - Cleaned, Outliers, NOT
Feature Engineered, all fields included.- (AllFeat)
And apply them across the datasets to create a comparison table of results I can use for the final model recommendations to promote to development of the demonstration model.
Conclusion and next steps
From the results of the model development and testing, two models will be selected and applied to the development of the interactive and live demonstrations
Execute
- Interpret model
- Evaluate model performance using metrics
- Prepare results, visualizations, and actionable steps to share with stakeholders
Conclusion
Project Close
- Under instruction from the client the following operations will be
carried
- Data Erasure - all supplied data will be backed up to client's cloud
- All local copies of data will be deleted for data security
- Agree Dates with client for follow up meetings at six and twelve months for HR and Team Leaders review meeting