I have variation models of data with different number of input data but same number of target data. The same data size of training, validation, testing were applied on those models. As we know, data size affect the accuracy/prediction. But what about using different number of input data, does it affect the accuracy/ prediction/ regression of the network too?
Dear Nur Asyikin Abd Jamel , yes the number of input variables(features) affect the regression results, so regression model which take a dependent variable and independent variables can be improved using feature engineering techniques. For example, you can find the importance of features, correlation analysis techniques to improve and optimize the model.
Imran Jamal Thank you for your answer. May I ask which one gives the best regression result? Using less number of input data, or more number of input data? As I would like to do further findings for the comparison and MSE of using a different number of input data
If you need to compare the performance results of regression model in term of MSE, RMSE, MAPE etc then for that you can make a separate table ..First find the correlation of the input with output data feature. drop the ones with similar correlation, this way you reduce the data features. Reducing the input data features using proper reduction or feature selection method can optimize the results of the regression model. For MAE sort of comparison train traditional and existing model on the same data and then compare the results in term of MAE, MAPE and RMSE.
- Badges
- Science topic
- Similar topics
- Organizational Psychology
- Training