A data analytics and machine learning project that predicts advertising-driven sales using regression models. This project highlights end-to-end skills in data analysis, feature engineering, modeling, and result interpretation.
This project uses historical data on advertising budgets across different channels to predict sales outcomes. The objective is to build reliable models that estimate how much revenue can be generated from advertising spend, thereby supporting data-driven marketing decisions.
It demonstrates the ability to take raw data, transform it, model it, and interpret predictions in a meaningful way — valuable for roles in analytics, marketing science, and data-driven decision making.
-
Data Exploration & Analysis
Visualization of relationships between ad spend (TV, radio, newspaper) and sales. -
Feature Engineering & Preprocessing
Data cleaning, scaling, transformations, and handling correlations. -
Model Training & Evaluation
Implementation and comparison of regression models (e.g. linear regression, regularized regression). Evaluation using metrics like RMSE, MAE, R². -
Model Serialization & Usage
Saving trained model(s) with methods for inference on new advertising budgets. -
Notebook / Experimentation
Jupyter notebook with exploratory data analysis, model experiments, and visualizations.
Advertising-Sales-prediction/
├── notebooks/ # contains EDA, modeling, experiments
├── data/ # dataset files (if applicable)
├── models/ # serialized model files
├── scripts/ # code for training, prediction
├── requirements.txt # project dependencies
└── README.md # this file
| Skill Area | Tools / Libraries |
|---|---|
| Data Analysis & Visualization | pandas, numpy, matplotlib, seaborn |
| Machine Learning & Regression | scikit-learn |
| Model Serialization | joblib / pickle |
| Experimentation | Jupyter Notebooks |
| Project Structuring | clean modular code, versioning |
Prerequisites
- Python 3.6+
- pip
Steps
- Clone the repository:
git clone https://github.com/UK183/Advertising-Sales-prediction.git cd Advertising-Sales-prediction - Install project dependencies:
pip install -r requirements.txt
- Run the notebook(s) to view EDA and model development:
jupyter notebook notebooks/Advertising_Sales_Prediction.ipynb
- Use the provided script to make predictions on new data (if script exists), e.g.:
python scripts/predict.py --tv 100 --radio 50 --newspaper 20
- Metrics such as RMSE, MAE, R² shown in notebook to compare model performance.
- Visualizations of predicted vs actual sales to assess model quality.
- Insights into which advertising channels contribute most to sales.
- Gained experience building regression models for real-world business applications.
- Learned feature engineering, handling multicollinearity, and model evaluation.
- Developed skills in packaging a model for inference and interpreting output into actionable insights.
Kazi Umar
Linkedin profile: https://www.linkedin.com/in/umar-kazi18
💼 Data Analyst | ML Engineer | Data Science & AI Enthusiast | Power BI | Python | SQL