Objective
Build a multi-class classification model that can predict a user's specific location (a unique room or space) within a building by using the Received Signal Strength Indicator (RSSI) from numerous nearby Wi-Fi Access Points.
Business Value
- Indoor Navigation Systems: Enable precise navigation in shopping malls, airports, hospitals, and large office buildings
- Location-Based Services: Provide targeted advertising, personalized recommendations, and context-aware mobile applications
- Asset Tracking: Track valuable equipment, inventory, and personnel within warehouses and industrial facilities
- Emergency Response: Quickly locate personnel during emergencies in large buildings
- Network Optimization: Identify critical Access Points for location services and optimize their placement
Core Libraries
- pandas & numpy: Data manipulation and numerical operations
- scikit-learn: Machine learning algorithms and evaluation metrics
- matplotlib: Data visualization and plotting
- RandomForestClassifier: Ensemble learning for multi-class location prediction
Dataset
Source: Kaggle - "UJIIndoorLoc Data Set"- Size: Wi-Fi fingerprints from 520 different Access Points
- Features: RSSI values from 520 WAPs (Wireless Access Points)
- Target: Building, floor, and room location combinations
- Scope: Multiple buildings with hundreds of unique indoor locations
- Quality: Comprehensive dataset with training and validation splits
Step-by-Step Guide
1. Environment Setup
# Install required packages
pip install pandas numpy scikit-learn matplotlib kaggle
# Set up Kaggle API for dataset download
# Upload kaggle.json file when prompted
2. Data Collection and Preprocessing
import pandas as pd
import numpy as np
from sklearn.preprocessing import LabelEncoder
# Download and load UJIIndoorLoc dataset
df_train = pd.read_csv('trainingData.csv')
df_val = pd.read_csv('validationData.csv')
df = pd.concat([df_train, df_val], ignore_index=True)
# Handle missing signal values (100 -> -105 dBm)
wap_cols = [f'WAP{str(i).zfill(3)}' for i in range(1, 521)]
df[wap_cols] = df[wap_cols].replace(100, -105)
# Create unique location identifier
df['location'] = (df['BUILDINGID'].astype(str) + '-' +
df['FLOOR'].astype(str) + '-' +
df['SPACEID'].astype(str))
3. Feature Engineering
# Extract Wi-Fi signal strength features
X = df[wap_cols] # 520 Access Point RSSI values
y = df['location'] # Unique location identifier
# Encode location labels
le = LabelEncoder()
y_encoded = le.fit_transform(y)
# Train-test split with stratification
from sklearn.model_selection import train_test_split
X_train, X_test, y_train, y_test = train_test_split(
X, y_encoded, test_size=0.3, random_state=42, stratify=y_encoded
)
4. Model Training
from sklearn.ensemble import RandomForestClassifier
# Configure RandomForest for high-dimensional problem
model = RandomForestClassifier(
n_estimators=50,
random_state=42,
n_jobs=-1 # Use all CPU cores
)
# Train the model
model.fit(X_train, y_train)
5. Model Evaluation
from sklearn.metrics import accuracy_score, classification_report
# Make predictions
y_pred = model.predict(X_test)
# Calculate accuracy
accuracy = accuracy_score(y_test, y_pred)
print(f"Overall Model Accuracy: {accuracy:.2%}")
# Detailed performance analysis
print(classification_report(y_test, y_pred))
6. Feature Importance Analysis
import matplotlib.pyplot as plt
# Identify most important Access Points
importances = model.feature_importances_
top_20_indices = np.argsort(importances)[-20:]
# Visualize critical APs for localization
plt.figure(figsize=(12, 10))
plt.barh(range(20), importances[top_20_indices])
plt.yticks(range(20), [f'WAP{i+1:03d}' for i in top_20_indices])
plt.title('Top 20 Most Important Access Points for Localization')
plt.xlabel('Feature Importance')
plt.show()
7. Real-World Deployment
# Function for real-time location prediction
def predict_location(rssi_readings):
\"\"\"
Predict indoor location from real-time RSSI readings
Args:
rssi_readings: Dict of AP_ID -> RSSI_value
Returns:
Predicted location string
\"\"\"
# Prepare feature vector
feature_vector = np.full(520, -105) # Default to no signal
for ap_id, rssi in rssi_readings.items():
if ap_id in wap_mapping:
feature_vector[wap_mapping[ap_id]] = rssi
# Predict location
prediction = model.predict([feature_vector])[0]
return le.inverse_transform([prediction])[0]
Success Criteria
- Primary Metric: Classification accuracy > 85% for room-level prediction
- Coverage: Successfully predict locations across multiple buildings and floors
- Robustness: Handle missing Access Point data gracefully
- Interpretability: Identify critical Access Points for network planning
- Speed: Real-time prediction capability for mobile applications
Next Steps & Extensions
Technical Enhancements
1. Advanced Algorithms: Experiment with XGBoost, neural networks, or ensemble methods
2. Feature Engineering: Include temporal patterns, device-specific calibration, and signal quality metrics
3. Dimensionality Reduction: Apply PCA or feature selection to reduce computational requirements
4. Calibration: Implement device-specific RSSI calibration for improved accuracy
Business Applications
1. Mobile App Integration: Develop SDKs for iOS/Android applications
2. Real-Time Systems: Build streaming prediction pipeline for live location services
3. Multi-Building Scaling: Extend to campus-wide or city-wide indoor positioning
4. Hybrid Positioning: Combine with BLE beacons, magnetometer, and accelerometer data
Research Directions
1. Transfer Learning: Adapt models trained on one building to work in another
2. Uncertainty Quantification: Provide confidence intervals for location predictions
3. Privacy Preservation: Implement federated learning approaches for sensitive locations
4. Energy Optimization: Balance prediction accuracy with mobile device battery consumption
Files in this Project
- README.md - Project documentation and implementation guide
- indoor_localization_wifi_rssi.ipynb - Complete Jupyter notebook implementation
- requirements.txt - Python package dependencies
Key Insights
- Wi-Fi RSSI fingerprints provide highly distinctive signatures for indoor locations
- RandomForest effectively handles the high-dimensional nature of 520 Access Point features
- Feature importance analysis reveals critical Access Points for network infrastructure planning
- Proper handling of missing signals (no-signal values) significantly improves model performance
- The approach scales well to hundreds of unique indoor locations with high accuracy