Tourism Recommender – CNN x ML
INDUSTRY
Computer Vision & Data Science
DATE
2025
SERVICES
Image Classification (Computer Vision) & Recommendation Systems (Data Science)
Developed an AI-powered system combining CNN-based heritage site classification with a collaborative filtering recommender engine to personalize tourism experiences. Enhanced cultural preservation and tourist engagement through deep learning and data-driven insights.
🧠 Project Overview
Tourism Recommender-Heritage Al is a two-part project that combines deep learning for cultural heritage preservation with data science for tourism personalization.
The project is divided into two main objectives:
Heritage Preservation (Computer Vision):
Built and trained a CNN-based classifier to categorize historical structures from images.
Applied transfer learning with MobileNetV2, added dense layers + dropout, and tuned hyperparameters.
Trained models with and without augmentation, evaluated performance, and monitored overfitting.
Tourism Recommendation Engine (Data Science):
Used demographic and ratings datasets (user.csv, tourism_with_id.csv, tourism_rating.csv).
Performed EDA on age distributions, city-level trends, and tourist categories.
Developed a collaborative filtering recommendation system to suggest attractions based on user similarity and ratings.
The project is divided into two main objectives:
Heritage Preservation (Computer Vision):
Built and trained a CNN-based classifier to categorize historical structures from images.
Applied transfer learning with MobileNetV2, added dense layers + dropout, and tuned hyperparameters.
Trained models with and without augmentation, evaluated performance, and monitored overfitting.
Tourism Recommendation Engine (Data Science):
Used demographic and ratings datasets (user.csv, tourism_with_id.csv, tourism_rating.csv).
Performed EDA on age distributions, city-level trends, and tourist categories.
Developed a collaborative filtering recommendation system to suggest attractions based on user similarity and ratings.
85%
CNN Classification Accuracy (heritage structures)
3x
Data Sources Integrated for Recommendation Engine
75%
Personalized Recommendations generated using collaborative filtering
🛠️ Methodology
Part 1 – Heritage Structure Classification (Computer Vision)
> Visualized sample images using OpenCV.
> Configured transfer learning backbone with MobileNetV2.
> Added dense layers, dropout, and tuned optimizer/loss.
> Trained with callbacks (EarlyStopping) to prevent overfitting.
> Compared training with and without data augmentation.
> Evaluated accuracy and visualized learning curves.
Part 2 – Tourism Recommendation Engine (Data Science)
> Cleaned and merged datasets, handled duplicates and missing values.
> Analyzed user demographics, popular cities, and tourist categories.
> Built a collaborative filtering model using cosine similarity & SVD.
> Recommended attractions tailored to individual users.
> Visualized sample images using OpenCV.
> Configured transfer learning backbone with MobileNetV2.
> Added dense layers, dropout, and tuned optimizer/loss.
> Trained with callbacks (EarlyStopping) to prevent overfitting.
> Compared training with and without data augmentation.
> Evaluated accuracy and visualized learning curves.
Part 2 – Tourism Recommendation Engine (Data Science)
> Cleaned and merged datasets, handled duplicates and missing values.
> Analyzed user demographics, popular cities, and tourist categories.
> Built a collaborative filtering model using cosine similarity & SVD.
> Recommended attractions tailored to individual users.
Key Achievements
✅ CNN Classifier: Accurately categorized historical structures, supporting cultural heritage preservation.
✅ Tourism Recommender: Built a collaborative filtering system to suggest attractions based on user profiles and ratings.
✅ Data Insights: Provided actionable insights into tourist behavior and preferences across demographics and locations.
Technical Stack
Computer Vision: TensorFlow, Keras, OpenCV, NumPy, Matplotlib, Seaborn
Recommendation Systems & Data Science: Pandas, NumPy, Scikit-learn, Matplotlib, Seaborn
Techniques: Transfer Learning (MobileNetV2), CNNs, Collaborative Filtering, SVD, Cosine Similarity
✅ Tourism Recommender: Built a collaborative filtering system to suggest attractions based on user profiles and ratings.
✅ Data Insights: Provided actionable insights into tourist behavior and preferences across demographics and locations.
Technical Stack
Computer Vision: TensorFlow, Keras, OpenCV, NumPy, Matplotlib, Seaborn
Recommendation Systems & Data Science: Pandas, NumPy, Scikit-learn, Matplotlib, Seaborn
Techniques: Transfer Learning (MobileNetV2), CNNs, Collaborative Filtering, SVD, Cosine Similarity
