Hotel Booking Demand Prediction

Hotel Booking Demand Prediction/Dataset/README.md

@@ -0,0 +1,7 @@
+# Hotel Booking Demand Dataset
+
+**Source**: [Hotel Booking Demand - Kaggle](https://www.kaggle.com/datasets/jessemostipak/hotel-booking-demand)
+
+**Description**:  
+This data set contains booking information for a city hotel and a resort hotel, and includes information such as when the booking was made, length of stay, the number of adults, children, and/or babies, and the number of available parking spaces, among other things.
+All personally identifying information has been removed from the data.

Hotel Booking Demand Prediction/Model/README.md

@@ -0,0 +1,102 @@
+# Hotel Booking Demand Prediction
+
+## 🎯 **Goal**
+
+Predicting cancellations based on booking data to estimate demand for hotel rooms.
+
+## 🧵 **Dataset**
+
+[Hotel Booking Demand Dataset](https://www.kaggle.com/datasets/jessemostipak/hotel-booking-demand)
+
+## 🧾 **Description**
+
+This data set contains booking information for a city hotel and a resort hotel, and includes information such as when the booking was made, length of stay, the number of adults, children, and/or babies, and the number of available parking spaces, among other things. The problem is binary classification of cancellation status to estimate hotel booking demand.
+
+## 🧮 **What I have done**
+
+1. Exploratory analysis of features: cleaning, preprocessing and data visualization.
+2. Feature engineering:
+    * re-categorizing categorical features based on target splits
+    * target-encoding high-cardinality categorical features
+    * discretizing numerical features with low number of unique values
+3. Feature selection:
+    * Statistical tests - Pearson correlation, Mutual information scores, ANOVA F-test, Chi-squared test of independence
+    * Model-based feature importances using Extremely-Randomized Trees.
+4. Created a holdout set for testing using Stratified sampling to maintain imbalance ratio.
+5. Training and validation of: Logistic Regression, Naive Bayes, K-nearest neighbours, Decision Tree, Random Forest, AdaBoost, Multi-Layer Perceptron, and gradient-boosting trees (XGBoost, CatBoost, LightGBM).
+6. Model ensembling using averaging of predictions with different configurations.
+7. Models were tuned and evaluated based on ROC-AUC score instead of Accuracy, since the target classes are imbalanced.
+
+## 🚀 **Models Implemented**
+
+* Logistic Regression
+* Naive Bayes: Gaussian
+* K-Nearest Neighbours
+* Decision Tree
+* Random Forest
+* AdaBoost
+* Neural network: Multi-layer Perceptron
+* Gradient-boosting models: XGBoost, CatBoost, LightGBM
+* Model Ensembling: Simple/Power/Weighted averaging
+
+## 📚 **Libraries Needed**
+
+* Pandas
+* Numpy
+* Scikit-learn
+* XGBoost
+* CatBoost
+* LightGBM
+* Matplotlib
+* Seaborn
+
+## 📊 **Exploratory Data Analysis Results**
+
+**Feature distributions**  
+![Image](../Images/featdist_leadtime.png)  
+![Image](../Images/featdist_arrivalweek.png)  
+![Image](../Images/featdist_arrivaldayofmonth.png)  
+![Image](../Images/featdist_staysweekend.png)  
+![Image](../Images/featdist_staysweekday.png)  
+![Image](../Images/featdist_totalstay.png)  
+![Image](../Images/featdist_adults.png)  
+![Image](../Images/featdist_adr.png)  
+
+**Feature selection**:  
+Correlation between features:  
+![Image](../Images/featselect_corrfeatures.png)  
+Correlation with target:  
+![Image](../Images/featselect_corrtarget.png)
+Mutual Information:  
+![Image](../Images/featselect_mutualinfo.png)  
+Model-based feature importances:  
+![Image](../Images/featselect_modelfimp.png)  
+
+## 📈 **Performance of the Models**
+
+Models were evaluated based on ROC-AUC score due imbalanced class ratio.
+
+| Model configuration | ROC-AUC Score  
+|:-----|:-----:
+| Logistic Regression | 0.8470  
+| Gaussian Naive Bayes | 0.7944  
+| K-Nearest Neighbours | 0.8810  
+| Decision Tree | 0.8820  
+| Random Forest | 0.8958  
+| AdaBoost | 0.8959  
+| Multi-layer Perceptron | 0.9039  
+| XGBoost | 0.9138  
+| LightGBM | 0.9146
+| CatBoost | 0.9154  
+| Simple averaging | 0.9108  
+| Power averaging | 0.9062  
+| **Weighted averaging** | **0.9159**  
+
+## 📢 **Conclusion**
+
+Trained a variety of models and created ensembles using averaging methods. Used ROC-AUC score to evaluate for imbalanced classification, and the best performance was shown by the Weighted-averaging ensemble.
+
+## ✒️ **Your Signature**
+
+Siddhant Tiwari  
+([Github](https://www.github.com/siddhant4ds) - [Kaggle](https://www.kaggle.com/sid4ds) - [LinkedIn](https://www.linkedin.com/in/siddhant-tiwari-ds/))

Hotel Booking Demand Prediction/README.md

@@ -0,0 +1,26 @@
+# Hotel Booking Demand Prediction
+
+## Project structure
+
+.  
+├── Dataset  
+│   ├── hotel_bookings.csv  
+│   └── README.md  
+├── Images  
+│   ├── featdist_adr.png  
+│   ├── featdist_adults.png  
+│   ├── featdist_arrivaldayofmonth.png  
+│   ├── featdist_arrivalweek.png  
+│   ├── featdist_leadtime.png  
+│   ├── featdist_staysweekday.png  
+│   ├── featdist_staysweekend.png  
+│   ├── featdist_totalstay.png  
+│   ├── featselect_corrfeatures.png  
+│   ├── featselect_corrtarget.png  
+│   ├── featselect_modelfimp.png  
+│   └── featselect_mutualinfo.png  
+├── Model  
+│   ├── eda_modeling_ensembling.ipynb  
+│   └── README.md  
+├── requirements.txt  
+└── README.md

Hotel Booking Demand Prediction/requirements.txt

@@ -0,0 +1,8 @@
+pandas==2.2.1
+numpy==1.26.4
+matplotlib==3.8.4
+seaborn==0.13.2
+scikit-learn==1.5.0
+xgboost==2.1.0
+catboost==1.2.5
+lightgbm==4.5.0