📊 TeleCom Data Science & ML Engineering — Beginner Journey

A beginner-friendly project exploring core Data Science and Machine Learning concepts using real-world TeleCom employee data. This project walks through the full data pipeline — from raw CSV ingestion to trained regression models.

---

🗂️ Project Structure

📁 Project Root
├── 📄 TeleCom.csv                          # Core dataset (salary & tax data)
├── 📄 CsvImport.py                         # EDA & data analysis
├── 📄 LinearRegression.py                  # Linear Regression model
│
├── 📁 PolynomialImportCsvTest/
│   ├── 📄 TeleComP.csv                     # Extended dataset
│   └── 📄 PolynomialRegression.py          # Polynomial Regression (script)
│
└── 📁 PolynomialPractice/
    ├── 📄 TeleComP.csv                     # Extended dataset (copy)
    ├── 📄 PolynomialRegressionExample.py   # Polynomial Regression (script)
    └── 📄 PolynomialTest.ipynb             # Polynomial Regression (notebook)

---

🧭 Learning Roadmap

Stage 1              Stage 2                  Stage 3
─────────────────    ──────────────────────   ──────────────────────────
📥 Data Import   →   📈 Linear Regression  →  📐 Polynomial Regression
& EDA                (Salary → TaxDeduction)   (Curved fit, degree=2)
CsvImport.py         LinearRegression.py        PolynomialRegression.py

---

🛠️ Tech Stack

Tool	Purpose
Python	Core programming language
pandas	Data loading, manipulation & groupby analysis
matplotlib	Data visualisation (bar charts, scatter plots)
scikit-learn	Machine learning models & evaluation metrics
numpy	Numerical operations & array handling
Jupyter Notebook	Interactive experimentation environment

---

Stage 1 — Data Import & Exploratory Data Analysis

File: CsvImport.py
Dataset: TeleCom.csv — 35 employees across departments with Salary and Tax Deduction data.

What this script does

• Loads the TeleCom CSV into a pandas DataFrame
• Performs descriptive statistics and group analysis
• Visualises employee distribution by department
• Calculates the correlation between Salary and Tax Deduction

Key Analyses Performed

# Total salary across all employees
total_score = df["Salary"].sum()

# Max and mean tax deduction
max_tax  = df["TaxDeduction"].max()
mean_tax = df["TaxDeduction"].mean()

# Gender breakdown — count, mean/max/min salary and tax
df.groupby('Gender')["Salary"].mean()

# Employee count per department
df.groupby("Department")["Gender"].count()

# Salary vs Tax Deduction correlation
corr = df["Salary"].corr(df["TaxDeduction"])

📊 Visualisation — Employee Count by Department

A bar chart is generated showing how many employees exist per department (Sales, Finance, ICT, Marketing, Consulting, Human Resource).

Output: Bar chart — "No. of Employees in Department"
X-axis: Department names
Y-axis: Employee count

📌 Key Finding

Correlation between Salary and Tax Deduction: ~0.97

A near-perfect positive correlation — as salary increases, tax deduction increases proportionally. This makes sense as the foundation for building a regression model.

---

Stage 2 — Linear Regression Model

File: LinearRegression.py
Goal: Predict TaxDeduction from Salary using a trained Linear Regression model.

Model Pipeline

Raw Data  →  Train/Test Split  →  Fit Model  →  Predict  →  Evaluate
(80% / 20%)    LinearRegression()   y_pred        R², MSE

Model Setup

X = data[['Salary']]       # Independent variable
y = data['TaxDeduction']   # Dependent variable

X_train, X_test, y_train, y_test = train_test_split(
    X, y, test_size=0.2, random_state=42
)

model = LinearRegression()
model.fit(X_train, y_train)
y_pred = model.predict(X_test)

📊 Evaluation Results

Metric	Value
R² Score	~0.94+
Mean Squared Error	Low (tight fit)
Regression Coefficient	Positive slope
Intercept	Base tax at zero salary

📊 Visualisation — Regression Line

A scatter plot is generated showing actual vs predicted tax deductions, with a red regression line overlaid.

Output: Scatter plot — "Linear Regression: Tax Deduction vs Salary"
Blue dots  →  Actual test data points
Red line   →  Predicted regression line

Training Set Statistics

Total employees   : 35
Training samples  : 28  (80%)
Test samples      : 7   (20%)

Max Salary (train) : ~R24,444
Min Salary (train) : ~R9,225
Mean Tax (train)   : ~R1,700

---

Stage 3 — Polynomial Regression

Files: PolynomialRegression.py / PolynomialTest.ipynb
Dataset: TeleComP.csv — Same employees with an extended NewTaxDeduction column.
Goal: Fit a curved (degree=2) polynomial model to capture any non-linear relationships.

Why Polynomial Regression?

Linear regression assumes a straight-line relationship. Polynomial regression adds squared terms so the model can capture curved patterns in data, which is useful when the relationship between features bends rather than stays perfectly linear.

Model Setup

from sklearn.preprocessing import PolynomialFeatures
from sklearn.linear_model import LinearRegression

poly = PolynomialFeatures(degree=2, include_bias=False)
x_poly = poly.fit_transform(x)

poly_reg = LinearRegression()
poly_reg.fit(x_poly, y)

📊 Visualisation — Polynomial Curve

A scatter plot with a smooth green polynomial curve is generated, showing how the model fits the data with a slight curve rather than a straight line.

Output: Scatter plot — "Polynomial Regression: Salary vs TaxDeduction"
Blue dots    →  Actual data points
Green curve  →  Fitted polynomial regression curve (degree=2)

---

💡 Key Concepts Learned

Concept	Where Applied
CSV loading with pandas	CsvImport.py
GroupBy & aggregation	CsvImport.py
Correlation analysis	CsvImport.py
Train/test split	LinearRegression.py
Model fitting & prediction	LinearRegression.py
R² and MSE evaluation	LinearRegression.py
Feature transformation	PolynomialRegression.py
Non-linear modelling	PolynomialRegression.py
Jupyter Notebook workflow	PolynomialTest.ipynb

---

🚀 How to Run

Prerequisites

python -m pip install pandas matplotlib scikit-learn numpy

Run each stage

# Stage 1 — EDA
python CsvImport.py

# Stage 2 — Linear Regression
python LinearRegression.py

# Stage 3 — Polynomial Regression
python PolynomialImportCsvTest/PolynomialRegression.py

# Or open the notebook
jupyter notebook PolynomialPractice/PolynomialTest.ipynb

---

📝 Notes

• This project uses a small dataset (35 employees) — ideal for learning core concepts without computational overhead.
• The high R² on linear regression (~0.94+) reflects the near-perfect real-world correlation between salary and tax deduction.
• Polynomial regression (degree=2) was explored as a next step to understand non-linear modelling, even though linear regression already performs well on this dataset.

---

Built as a beginner introduction to Data Science and ML Engineering in Python.