Dear my friends.
I want to know the possibles ways for combining predictors from forecasting models (for example, combine different ANN models) in order to achieve a better point forecasting model, or even combine models from a population of an evolutionary algorithm.
Is there any technique for combining them optimally? For example, using some mathematical programming model, fitting the best parameters of a linear regression?
What are other options for combining these models?
Dear Vitor,
In linear regression you need one continuous dependent variable (e.g. depression)
and two or more continuous independent variables e.g. social support and aggression (you can also use dichotomous independent variables e.g. male, female. )
What your parameters are and what they mesures depend on what kind of research you are doing
MP finds the max/min values of objective function.
So, I developed the linear discriminant function, the object of which is to minimize the error rate. I think my several papers after 2014 help your approach.
You can try the AdaBoost technique, such as in
https://www.researchgate.net/publication/270397810_Comparison_of_four_Adaboost_algorithm_based_artificial_neural_networks_in_wind_speed_predictions
You can also apply Noise Peeling techniques to improve the AdaBoost results further.
https://www.researchgate.net/publication/281464742_Noise_Peeling_Methods_to_Improve_Boosting_Algorithms
Article Comparison of four Adaboost algorithm based artificial neura...
Article Noise Peeling Methods to Improve Boosting Algorithms
- Badges
- Science topic
- Similar topics
- Philosophy Of Science
- Models