I have a dataset from which I identified the set of features. Many existing work test this data set with multiple classification models including Decision Tree, Ensemble classifiers (Random Forest, XGBoost, GBDT). I want to test this data with new classification model, that can perform better than these models, so that I can add new contribution in my paper.
Have you tried Deep Learning Algorithms ?! or, Extreme Learning Machine (ELM) ?
There are many types of ANN have been proposed and used for classification problems, you may check them. (CNN, RNN, DBN .... ).
Regards,
Md. Razu Ahmed My problem is trajectory data mining, transportation mode classification. this is supervised learning problem.
Actually machine learning models can perform differently with respect to the data modality. For instance, Ensemble methods perform well with sensor-stream data or time-series data but they are not efficient when dealing with image data. Currently, deep learning networks have been promising. If your data is non-temporal and time-series, you can use 1-D convolution neural networks, please refer to (Deep-Sense Framework). If you are dealing with cross-domain try bi-view networks. If you are dealing with limited amount of data use active learning approach with any non-linear classifier such as ELM i would suggest. Lastly, if you are dealing with NLP kind of data, use topic modeling or LDA.
- Badges
- Science topic