Publicação
Machine learning approach for personalized recommendations on online platforms: uniplaces case study
| Resumo: | The goal of this project is to develop a model to personalize the user recommendations of an online marketplace named Uniplaces. This online business offers properties for medium and long-term stays, where landlords can directly rent their place to customers (mainly students). Whenever a student makes a reservation, the booking must be approved by the property owner. The current acceptance rate is 25%. The model is a response to this low acceptance rate, and it will have to show to each student the properties that are more likely to be accepted by the landlord. As a secondary objective, the model seeks to identify the reasons behind the landlord’s decision to accept or reject bookings. The model will be constructed using information from the users, landlord and the property itself kindly provided by Uniplaces. This information will pre-process with data cleaning, transformation and features reduction (where two techniques were applied: dimensionality reduction, features selection). After the data processing, several models were applied to the normalized data. The predictive models that will be applied are already being used on other online markets and platforms like Airbnb, Netflix or LinkedIn, namely Support Vector Machine, Neural Networks, Decision Tree, Logistic Regression and Gradient Boosting. The probability of acceptance proved to be very easy to predict, all the models predict 100% of the test dataset when using the Principal Component Analysis as the Dimensionality Reduction technique. This can be explained mainly by the fact that the new calculated features have a strong correlation with the target variable. All the algorithms predict 100% of the target variable when using Principal Component Analysis as a technique of dimensionality reduction. |
|---|---|
| Autores principais: | Villar, Marta Maria Cabral Menéres Posser |
| Assunto: | Machine Learning Data Mining Predictive Models Supervised Learning |
| Ano: | 2021 |
| País: | Portugal |
| Tipo de documento: | dissertação de mestrado |
| Tipo de acesso: | acesso aberto |
| Instituição associada: | Universidade Nova de Lisboa |
| Idioma: | inglês |
| Origem: | Repositório Institucional da UNL |
| Resumo: | The goal of this project is to develop a model to personalize the user recommendations of an online marketplace named Uniplaces. This online business offers properties for medium and long-term stays, where landlords can directly rent their place to customers (mainly students). Whenever a student makes a reservation, the booking must be approved by the property owner. The current acceptance rate is 25%. The model is a response to this low acceptance rate, and it will have to show to each student the properties that are more likely to be accepted by the landlord. As a secondary objective, the model seeks to identify the reasons behind the landlord’s decision to accept or reject bookings. The model will be constructed using information from the users, landlord and the property itself kindly provided by Uniplaces. This information will pre-process with data cleaning, transformation and features reduction (where two techniques were applied: dimensionality reduction, features selection). After the data processing, several models were applied to the normalized data. The predictive models that will be applied are already being used on other online markets and platforms like Airbnb, Netflix or LinkedIn, namely Support Vector Machine, Neural Networks, Decision Tree, Logistic Regression and Gradient Boosting. The probability of acceptance proved to be very easy to predict, all the models predict 100% of the test dataset when using the Principal Component Analysis as the Dimensionality Reduction technique. This can be explained mainly by the fact that the new calculated features have a strong correlation with the target variable. All the algorithms predict 100% of the target variable when using Principal Component Analysis as a technique of dimensionality reduction. |
|---|