Project 01: Instagram Reach Prediction - Discovery Challenge

🎯 Learning Objectives

By completing this project, you will:

• Master Data Exploration: Explore and understand real-world Instagram engagement data
• Apply Data Preparation Techniques: Clean data, handle missing values, and scale features
• Build a Regression Model: Create a Linear Regression model to predict engagement
• Evaluate Model Performance: Use metrics like R² and coefficients to assess your model
• Make Real Predictions: Predict engagement for new Instagram posts

Difficulty Level: Intermediate Estimated Time: 2-3 hours Prerequisites: Completed Activity 05 (Data Preparation) and Activity 06 (Regression)

🚀 Getting Started (See Results in 30 Seconds!)

Option One: Explore the Complete Solution First (Recommended)

1. Open v1-baseline-100percent.ipynb in Google Colab or Jupyter
2. Run all cells (Runtime -> Run all)
3. Observe the complete workflow from data loading to predictions
4. Study the pattern: Import -> Explore -> Prepare -> Train -> Predict

Option 2: Jump Straight to the Challenge

1. Open project-01-instagram-reach.ipynb
2. Run the setup cells to see working examples
3. Complete TODOs in order to build your own predictor

📋 What's Already Working

✅ Data Download & Loading

• Dataset download code (Instagram.csv from Google Drive)
• Data loading with proper encoding (latin1)
• All necessary library imports

✅ Data Exploration (Phase 2)

• Code to view first 5 rows
• Dataset info and shape
• Missing value detection
• Word cloud visualizations for captions and hashtags
• Correlation analysis

✅ Data Preparation Setup

• Feature selection (Likes, Comments, Shares, Profile Visits, Follows)
• MinMaxScaler import and initialization
• Train-test split setup code

✅ Model Evaluation (Bonus)

• Complete code for MAE, MSE, RMSE calculations
• User interface for predictions (Extension Challenge)

📝 Your Tasks (5 TODOs to Complete)

Phase 3: Data Preparation (2 TODOs - Medium)

Success Criteria:

• x_scaled values should be between 0 and 1
• x_train.shape should be (66, 5)
• x_test.shape should be (33, 5)

Phase 4: Model Building (3 TODOs - Medium to Hard)

Success Criteria:

• Model accuracy should be approximately 0.8461 (84.61%)
• Coefficients should show 5 values (one per feature)
• Custom prediction should return ~9300-9500

✅ Success Criteria

You've successfully completed this project when:

• All 5 TODO cells execute without errors
• MinMaxScaler properly scales features to 0-1 range
• Train-test split produces correct shapes (66 train, 33 test samples)
• Model accuracy (R²) is above 80%
• Model coefficients and intercept are printed
• Custom prediction returns a reasonable engagement value

🎓 Key Concepts Applied

One. Feature Scaling

from sklearn.preprocessing import MinMaxScaler
scaler = MinMaxScaler()
scaler.fit(x)
x_scaled = scaler.transform(x)

Why? Machine learning algorithms work better when features are on the same scale.

2. Train-Test Split

from sklearn.model_selection import train_test_split
x_train, x_test, y_train, y_test = train_test_split(
    x_scaled, y, test_size=0.33, random_state=42
)

Why? We need separate data to train and evaluate our model fairly.

3. Linear Regression

from sklearn.linear_model import LinearRegression
model = LinearRegression()
model.fit(x_train, y_train)
accuracy = model.score(x_test, y_test)

Why? Linear regression finds the best linear relationship between features and target.

🚀 Extension Challenges

Once you complete all TODOs, try these:

1. User Interface

   • Create an interactive predictor using input() prompts
   • Display engagement level categories (Viral, Great, Keep creating)
   • Full code provided in Extension section

2. Model Evaluation Deep Dive

   • Calculate MAE, MSE, and RMSE metrics
   • Compare predictions vs actual values
   • Code provided in notebook

3. Experiment with Features

   • Try removing different features and see how accuracy changes
   • Which features have the biggest impact on engagement?

4. Visualize Results

   • Create scatter plots comparing predictions vs actual values
   • Plot residuals to understand model errors

💡 Tips for Success

1. Run v1 baseline first: See the complete solution to understand the workflow
2. Follow the phases: Data Prep -> Model Building -> Evaluation
3. Check shapes: Use .shape to verify data dimensions
4. Print intermediate results: Add print() statements for debugging
5. Remember to scale: Always scale test data with the fitted scaler

🐛 Common Issues & Solutions

Issue: "ValueError: Expected 2D array"

Solution: Wrap prediction input in double brackets: [[282.0, 4.0, 9.0, 165.0, 54.0]]

Issue: "NameError: scaler is not defined"

Solution: Run the cell that creates the MinMaxScaler first

Issue: "Different shapes for train and test"

Solution: Verify you're using x_scaled (not x) in train_test_split

Issue: "Accuracy is very low or negative"

Solution: Check that you scaled the data BEFORE splitting, and used the correct target variable (y)

Issue: "Predictions seem off scale"

Solution: Make sure to scale the prediction input using scaler.transform() before model.predict()

📊 Expected Outputs

Phase 3 - Data Shapes

x_train shape: (66, 5)
x_test shape: (33, 5)
y_train shape: (66,)
y_test shape: (33,)

Phase 4 - Model Results

Model Accuracy: 0.8461
Model coefficients: [115.67681478 1898.72802154 -394.46756815 -495.96781512 748.19781634]
Model intercept: 2358.9711885315516

Custom Prediction

For input [282.0, 4.0, 9.0, 165.0, 54.0]:

Predicted engagement: ~9300-9500

📚 Additional Resources

• sklearn LinearRegression Docs
• Understanding MinMaxScaler
• Instagram Analytics Guide

🎯 Estimated Completion Time

• Fast Track (completed Activity 06): 60-90 minutes
• Standard Pace: 90-120 minutes
• Deep Learning (with extensions): 150-180 minutes

Ready to begin? Open project-01-instagram-reach.ipynb and start building your Instagram engagement predictor! 🚀