This project is developed to predict the rainfall in millimeters in Ramannapeta of Yadadri-Bhuvanagiri district of Telangana.
This data set is real time dataset and we have done from data pre-processing to metrics evaluation
Dineshwar Doddapaneni
P.Lakshmi manaswini (https://github.com/LakshmiManaswini-7)
Data is collected from open data websites
Here we used 3 different methods to predict the rainfall,
1)Deep Neural Networks
2)Decision Tree Regressor
3)Random Forest Regressor
Here we have used a deep neural network of 4 hidden layers,
After training the data seeing the MSE we thought this were not the best choice(41 in training)
- We havent managed to get high quantity of data
- we have trained the machine only with 1000 epochs so due to lack of computational power
- The architecture was also not sufficent
Here we used max depth equal to 5
the R2 score for this model 0.96(train)
the mean absolute error is 0.45(train) 6.9(test)
here we used Random forest Regressor
the R2 score for this model 0.86(train)
the mean absolute error is 1.029(train) 4.8(test)
- out this we felt Decision Tree Regressor algorithm is best
- Here we can collect more data and use neurals networks
- more computational power could be really usefull for us