Multiple Linear Regression

Nerd Cafe

Module 1: Introduction to Multiple Linear Regression

What is Multiple Linear Regression?

• A supervised learning algorithm.

• Predicts a dependent variable (target) using multiple independent variables (predictors).

• Extension of simple linear regression.

Mathematical Representation:

$$\hat{y}=\omega_{0}+\omega_{1}x_{1}+\omega_{2}x_{2}+...+\omega_{p}x_{p}$$

Where:

• $\hat{y}$​: predicted value

• $\omega_{0}$​: intercept

• $\omega_{1} , ... \omega_{p}$​: coefficients

• $x_{1} , ... , x_{p}$​: independent variables

Module 2: Assumptions of Multiple Linear Regression

• Linearity: Target vs predictors must be linear.

• Independence: Observations are independent.

• Homoscedasticity: Equal variance of residuals.

• Normality: Residuals follow a normal distribution.

• No Multicollinearity: Predictors are not highly correlated.

• No Autocorrelation: Residuals are independent.

• Fixed Independent Variables: Same across all samples.

Module 3: Python Implementation (Step-by-Step)

Step 1: Data Preparation

I am using below dataset:

1KB

datamlr.csv.csv

Open

1. Importing Libraries

import numpy as np
import pandas as pd
import matplotlib.pyplot as plt

• numpy (np): Used for numerical operations (e.g., arrays, linear algebra).

• pandas (pd): Used for data manipulation and analysis using DataFrames.

• matplotlib.pyplot (plt): Used for plotting graphs and visualizations.

2. Load Dataset

# Load data
dataset = pd.read_csv('Dataset/datamlr.csv')

• This line loads a CSV file named db_mlr.csv into a DataFrame called dataset.

• dataset now holds tabular data—rows and columns like an Excel sheet.

3. Feature Matrix (X) and Target Vector (y)

X = dataset[['R&D Spend', 'Administration', 'Marketing Spend']]
y = dataset['Profit']

• X is a DataFrame containing the input features:

  • 'R&D Spend'

  • 'Administration'

  • 'Marketing Spend'

• y is a Series containing the target variable (what you want to predict), which is 'Profit'.

Step 2: Splitting the Dataset

1. Importing the Function

from sklearn.model_selection import train_test_split

• This imports the train_test_split function from Scikit-learn, a popular machine learning library in Python.

• train_test_split() is used to split your dataset into two parts:

  • Training set (to train the model)

  • Testing set (to evaluate how well the model performs)

2. Splitting the Data

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

Here's what it does:

• X : Your features (independent variables) — e.g., R&D Spend, Administration, Marketing Spend

• y : Your target/output — e.g., Profit

It returns four variables:

1. X_train — 80% of the data for training (features)

2. X_test — 20% of the data for testing (features)

3. y_train — 80% of the target values for training

4. y_test — 20% of the target values for testing

Parameter:

• test_size=0.2 means:

  • 20% of the data will go into the test set

  • 80% will go into the training set

Step 3: Model Training

from sklearn.linear_model import LinearRegression

• This line imports the LinearRegression class from the sklearn.linear_model module.

• LinearRegression is a machine learning model that assumes a linear relationship between input features (X) and the output (y).

regressor = LinearRegression()

• This line creates an instance of the LinearRegression class named regressor.

• At this stage, the model is initialized but not trained yet.

regressor.fit(X_train, y_train)

• This line trains (fits) the linear regression model using training data.

• X_train is the matrix of input features for training.

• y_train is the corresponding target/output values.

• The .fit() method calculates the best-fit line (or hyperplane) that minimizes the mean squared error between the predicted and actual y values.

Step 4: Model Testing and Prediction

y_pred = regressor.predict(X_test)

• This line uses the trained linear regression model (regressor) to make predictions on the test data (X_test).

• y_pred will contain the predicted output values corresponding to the input features in X_test.

# Compare actual vs predicted
df = pd.DataFrame({'Real Values': y_test, 'Predicted Values': y_pred})

• This line creates a DataFrame using pandas (assumes import pandas as pd has been done).

• The DataFrame df shows a side-by-side comparison of:

  • 'Real Values': The actual output values (y_test) from the test set.

  • 'Predicted Values': The predicted values (y_pred) from the model.

print(df)

This line simply prints the DataFrame to the console, allowing you to visually compare the actual and predicted results.

    Real Values  Predicted Values
27     103282.4     111763.191880
18     124266.9     128057.177230
11     144259.4     136245.503634
16     126992.9     115683.954593
8      152211.8     149899.284176
26     105733.5     109892.305745

Step 5: Model Evaluation

1. Imports:

from sklearn.metrics import mean_squared_error, mean_absolute_error, r2_score
from math import sqrt

• mean_squared_error: Calculates the average of squared differences between actual (y_test) and predicted (y_pred) values.

• mean_absolute_error: Calculates the average of absolute differences.

• r2_score: Calculates the R-squared (coefficient of determination), which indicates how well the model explains the variance in the data.

• sqrt: From Python’s math module, used to compute the square root (for RMSE).

2. Metric Calculations:

mse = mean_squared_error(y_test, y_pred)

• MSE (Mean Squared Error): Measures the average of the squared errors.

• A lower MSE indicates better model performance.

rmse = sqrt(mse)

• RMSE (Root Mean Squared Error): The square root of MSE, which brings the error back to the original unit of the target variable.

• Easier to interpret than MSE.

mae = mean_absolute_error(y_test, y_pred)

• MAE (Mean Absolute Error): Measures the average magnitude of the errors (without squaring).

• It’s more robust to outliers than MSE.

r2 = r2_score(y_test, y_pred)

• R² Score: Indicates the proportion of variance in the target variable that is predictable from the features.

• Ranges from:

  • 1.0 (perfect prediction)

  • 0 (no explanatory power)

  • < 0 (worse than baseline model)

3. Output

MSE: 50174701.37246796
RMSE: 7083.410292540449
MAE: 6344.205408672026
R²: 0.8457097502712363

Step 6: Making Predictions for New Data

new_data = [[166343.2, 136787.8, 461724.1]]
profit = regressor.predict(new_data)
print(profit)

Output

[191057.03223215]

Step 7: Interpreting Model Coefficients

print("Coefficients:", regressor.coef_)
print("Intercept:", regressor.intercept_)

Output

Coefficients: [ 0.82965119 -0.06684259  0.01573258]
Intercept: 54929.336292630556

Module 4: Applications

Field	Application Example
Finance	Stock price prediction
Marketing	Campaign effectiveness
Real Estate	House price prediction
Healthcare	Treatment outcome prediction
Economics	GDP or inflation forecasting
Social Sci.	Election result modeling

Module 5: Common Challenges

Challenge	Description
Multicollinearity	Highly correlated predictors affect model stability
Overfitting	Model too closely fits training data
Underfitting	Model too simple to capture patterns
Outliers	Affect predictions drastically
Missing Data	Leads to biased results
Non-Linearity	Linear model may not fit non-linear relationships

Module 6: Simple vs Multiple Linear Regression

Feature	Simple Linear Regression	Multiple Linear Regression
Independent Vars	One	Two or more
Complexity	Low	Higher
Use Case Example	Sales vs Ad Spend	Sales vs Ad, Price, Competition

Keywords

multiple linear regression, machine learning, regression model, independent variables, dependent variable, R&D spend, administration, marketing spend, profit prediction, data preprocessing, train test split, sklearn, linear regression, model evaluation, mean squared error, root mean squared error, R-squared score, model coefficients, multicollinearity, residual error