Abstract: The project is based on Regression and analysing the data. We in-take the data in matrix form and then train the parameters so that they can fit to predict the housing prices/rents. The Code goes through multivariate polynomial regression and uses basic approach of machine learning.
Tools Used: • Octave/MATLAB/jupyter/python(packages like numPy,pandas) • Optimization Function and their libraries ex. Conjugate gradient and F-min Unconstrained, feature scaling.
Approach: The problem of predicting housing prices depends on various factors like size of the house, no. Of rooms, size of family purchasing it, availability of hospitals and schools nearby, distance from main market, ZIP code, PIN code, City, Locality, Average income of neighbourhood, lifestyle grading of area, Crimes in the area and many more.
Now each of the factors has its own priority from the point of view of different buyers. A family with children might like the availability of school and some with elders might prefer availability of hospital. So there is inter-dependency among the attributes hence we cannot treat them independently and to use this, we have used quadratic (polynomial) degrees of attributes so that their inter-dependency can be treated and also one attribute might have priorities over the others so it’s better to treat them in different levels.
The pseudo code of the algorithm is given below:-
- Fetch the data from CSV file (or excel file).
- Separate needed data and their definition from the file and mould them into a proper matrix.
- Apply mapping of attributes to achieve interconnection and quadratic features.
- Train the parameters to compute cost and reduce it to an extent possible.
- Check the accuracy with 30% of the data as test data.
- Predict the test cases for various new inputs and find the accuracy of prediction.
Log in or sign up for Devpost to join the conversation.