1. I have developed a probabilistic classifier whose prediction is an output PDF.
2. I have multiple measure signals M1, M2, M3 belonging to the same true object X
3. Using each measured signal M1, M2 --Mn predictions P1, P2,-- Pn
I want to know what are the ways by which I can combine predictions of P1, P2,-- Pn to have better accuracy of detecting X?
Derar Varun Garg ,
Prediction methodology is a set of techniques used for forecasting the future. Futurology used such techniques as linear projections and extrapolations from trends, scenario-building, and what-if stories.
https://en.wikibooks.org/wiki/Future/Prediction_Methods
https://machinelearningmastery.com/combine-predictions-for-ensemble-learning/
Regards,
Shafagat
Hi Varun,
You may either combine the predictions through a Gaussian mixture model.
I think you are talking about Ensemble learning. In that case, you may use bagging or boosting strategies or voting methods:
Or you may also apply model selection to select the best performing model: https://machinelearningmastery.com/probabilistic-model-selection-measures/.
Hope this helps!
hi, using results of predict_proba method (from ml-model for example in sklearn-library) you can differentiate your probabilities other words you can define probabilities range of concrete object
- Badges
- Science topic