The Hotel Dilemma

• UW FinTech Boot Camp - Project 3 Submission - May 2021

Team Members

• Monique T.
• Tony H.
• April A.
• Nick N.

Motivation

• After a 1-year pandemic, we all need a vacation...
• Our group wanted to conduct a project using the tools learned in both Module 1 & Module 2 of the boot camp (python + machine learning). We set out looking for data sets and questions related to supply chain management and demand forecasting, which led us to a very interesting prompt related to forecasting hotel reservation cancellations. Given we all need a good vacation, we pursued the the following project...

Research Questions

• Using data and machine learning models, can hotel reservation cancellations be predicted?
• If yes to the above question, what models and methods most accurately predict hotel reservation cancellations?

Objectives

• Build a ML model that can predict whether a hotel reservation will be cancelled
• Analyze and understand data via organization, visualization, and dashboards

Data Sources

• Hotel Booking Demand

Action Items

• Data Cleaning & Shaping
  • Data comes from/affiliated with an article: Hotel Booking Demand Datasets
  • Data was cleaned by Thomas Mock and Antoine Bichat (additional cleaning and shaping conducted by our team)
  • Is there further noise/info we want to weed out? Label encoding?
• Machine Learning Model
  • Which model to use (try multiple models)
  • Ensemble/Classifier/Decision Tree/Regression? Pick several and also apply resampling techniques if needed. We predict classifier models will be the most effective given we will be classifying a binary outcome (cancelled vs not cancelled)
  • Which parameters/inputs produce the best outcomes (train/test split; different inputs for each ML model type aka reference documentation; which models are most efficient; what features in the dataset can we eliminate)
  • Look at data in different ways? Is the model/data better for predicting in the summer/winter/fall/etc? Should we try forecasting for specific date ranges, like spring break, holiday breaks, etc. This will be a reach if we have time.
• Data Visualization
  • Visualize by city hotel & resort hotel
  • Visualize by season
  • Visualize different demographics
  • Visualize different ML model outcomes?
  • Explore other means and methods of visualization that may give unique insight into the data set

Work Assignments

• ML Models & Workbook - April + Nick
• Data Visualizations & Dashboard - Monique + Tony
• Slide Show - Whole Team

Technologies

• Jupyter lab
• Python
• Pandas
• Numpy
• Sklearn
• Pyviz
• More to be imported and utilized in our python files

Attachments

• Analysis Folder - ML python files
• Visualizations Folder - visualization python files
• Data Folder- original data set
• Final presentation deck

Outcomes

• Using machine learning models, our Team was able to predict hotel cancellations with confidence (particularly using the SMOTEENN Resampling + BalancedRandomForestClassifier model, which rendered a ~90% accuracy score). Vizualizations of our final accuracy score outcomes are below. Please see our uploaded slide show for more information on the outcomes.

Presentation Assignments

• Intro & Hypothesis: Tony
• Visualizations & Intro to Data: Monique
• Data Preparation & Model Selection: Nick
• Model Outcomes & Takeaways: April