HabitatHive_oneAPI
Sudhanthiran N
Unknown
The project aims to provide properties based on the Habitability Index for people finding properties to rent. This Habitability index is found by an ML model that's backed by oneAPI. ...learn more
Project status: Under Development
oneAPI, Artificial Intelligence
Overview / Usage
The objective of this project is to develop a machine learning model that can recommend suitable properties for individuals who are relocating to a new city or location. To achieve this goal, a comprehensive dataset comprising of diverse properties located throughout the United States is collected and subjected to a series of preprocessing and cleansing steps. The processed data is then visualized to enhance the understanding of the dataset. The machine learning model receives input from the user regarding the property type and location preferences. Subsequently, based on the user’s priorities, the model identifies and extracts the top five properties with the highest habitability scores. The habitability score is calculated using the oneAPI toolkit and takes into consideration essential end-user needs such as power backup, traffic density level, and water supply.
Methodology / Approach
The project utilizes a dataset obtained from Kaggle, comprising of various features such as Property ID, Property type, property area, Number of windows, Number of doors, Furnishing, Frequency of Powercuts, Power backup, Water supply, traffic density score, crime rate, dust and noise, air quality index, neighbourhood review, and habitability score. As the habitability score prediction requires numerical values, all the discrete values were converted into numerical values (e.g., furnished, semi-furnished, and unfurnished to 1, 0.5, and 0, respectively). To simplify the dataset, a new dataset was created only comprising common values such as property area, frequency of power cuts, power backup, water supply, traffic density score, crime rate, dust and noise, air quality index, and neighbourhood review for a particular area. This preprocessed dataset will be used for further prediction and model training.
Technologies Used
The project utilized two machine learning models, namely the DecisionTreeRegressor and AdaBoostRegressor. The DecisionTreeRegressor was chosen as the weak learner, and it was fitted inside the AdaBoostRegressor. AdaBoostRegressor is a meta-estimator that adjusts the weights of instances based on the error of the current prediction. This approach is less prone to overfitting as the input parameters are not jointly optimized, and it improves the accuracy of weak classifiers.
The model training process was carried out using oneAPI, which has advantages such as improved accuracy and efficiency. The accuracy of the model trained using Google colab is approximately 89.5 to 90 out of 100, while the accuracy of the model trained with oneAPI is around 90.5 to 91.0. This improved accuracy is attributed to the use of oneAPI, which facilitated the development of a more precise model.
Documents and Presentations
Repository
https://github.com/fosslover69/habitat-hive