[WEEK 5] Prediction Of Real Estate Price
Team Members: Batuhan Ündar , Enes Koçak , Muhammed İkbal Arslan
This week, as we planned before we started to combine our model with CNN architecture. Firstly we will train 3 convolutional neural network. Then we will give the results from CNN as a parameter to the artificial neural network with other continuous parameters. Our goal in doing this is to use the picture data as a continuous parameter when making a price estimate.
How will it work?
Firstly every CNN will return us the quality score in return for the given image. The quality score mentioned here will be between 1 and 10. The quality scores in train data are given on the basis of prices when training the network. In this way, our system will be able to return the quality score according to the photos and will be effective for the home price forecast. at least hope it does :)
Why should we use three different CNN's, this is because we have different pictures like bathroom, kitchen and bedroom. You can think that we need to get the same quality score for each room according to the price indexed education. But remember that these points are not a result, these are parameters. That is, we are waiting for the artificial neural network to know which room pictures are effective on the price.
As the training process can take too long, we will get support from our university and start testing after getting our trained models. CNN will remain on hold until the training is completed and we will continue to improve our previous work.
