https://github.com/yaswanth1702/salary-prediction-in-the-ai-job-market

https://github.com/yaswanth1702/salary-prediction-in-the-ai-job-market

feature-engineering linear-regression model-interpretation model-tuning non-linear-regression

Last synced: 3 months ago
JSON representation

• Host: GitHub
• URL: https://github.com/yaswanth1702/salary-prediction-in-the-ai-job-market
• Owner: Yaswanth1702
• License: mit
• Created: 2025-01-28T08:49:48.000Z (5 months ago)
• Default Branch: main
• Last Pushed: 2025-01-28T09:00:15.000Z (5 months ago)
• Last Synced: 2025-01-28T09:29:46.865Z (5 months ago)
• Topics: feature-engineering, linear-regression, model-interpretation, model-tuning, non-linear-regression
• Language: R
• Homepage:
• Size: 0 Bytes
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# AI Job Market Salary Prediction

## Overview

This R project is focused on predicting AI-related job salaries using various machine learning models. The project leverages the `ai_job_market_insights.csv` dataset, which includes features like job title, industry, experience level, and location to predict salaries. The goal is to experiment with multiple models, including linear and non-linear approaches, to identify the most accurate salary prediction model.

## Key Features

- **Data Preprocessing**: 

  - Missing values handling using `na.omit()`

  - Box-Cox transformation to normalize the salary distribution

  - Dummy variable creation for categorical variables

  - Near-zero variance feature removal

  - Centering and scaling of numeric features for model training

- **Modeling**:

  - **Linear Models**: Linear Regression, Ridge, Lasso, and Elastic Net.

  - **Non-linear Models**: Neural Networks, MARS (Multivariate Adaptive Regression Splines), Support Vector Machines (SVM), and K-Nearest Neighbors (KNN).

  

- **Evaluation**:

  - Model evaluation metrics: RMSE (Root Mean Squared Error), R-squared, MAE (Mean Absolute Error).

  - Feature importance visualizations using `varImp()`.

- **Visualization**:

  - Visualize the distribution of salary before and after Box-Cox transformation.

  - Compare model performance using various metrics.

## Dataset

**Categorical Variables:**

- Job Title: Titles associated with various AI roles (e.g., Data Scientist, AI Engineer).

- Industry: The sectors employing AI professionals (e.g., Tech, Healthcare).

- Experience Level: Categorization based on the required experience (e.g., Entry-Level, Mid-Level).

- Location: Geographic region of employment.

- Education: Minimum educational qualifications (e.g., Bachelor’s, Master’s).

- Company Size: Size of the employing company (e.g., Small, Medium, Large).

- Remote Work: Proportion of work performed remotely.

- Job Posting Date: Period when the job was posted.

- Hiring Status: Current hiring stage (e.g., Open, Closed).

  

**Target Variable:**

- Salary (USD): Annual compensation for the AI roles.

## Required Libraries

- `dplyr`

- `caret`

- `ggplot2`

- `glmnet`

- `earth`

- `kernlab`

- `nnet`

- `gridExtra`

- `e1071`

To install these libraries, run the following R command:

```R

install.packages(c('dplyr', 'caret', 'ggplot2', 'glmnet', 'earth', 'kernlab', 'nnet', 'gridExtra', 'e1071'))

```

## Results and Evaluation

- **Linear Models**:

  - Linear Regression, Ridge Regression, Lasso Regression, and Elastic Net are trained and evaluated.

  

- **Non-linear Models**:

  - Neural Networks, MARS, SVM, and KNN are also trained and evaluated.

  

- **Model Performance**:

  - The script outputs evaluation metrics like RMSE, R-squared, and MAE for each model.

- **Best Model**:

  - The model performance metrics will indicate which model performs best for salary prediction.

## Future Work

- Experiment with ensemble models such as Random Forest or XGBoost.

- Perform additional feature engineering to improve model accuracy.

- Try hyperparameter tuning to optimize model performance further.