NPSC is a Matlab software library for online stream classification. It aims to provide flexible modeling, learning, and inference of general dependent nonparametric models for classifying stream data with non-stationary generating process to capture the concept drift occur in large volume of various stream data arising from social networks, financial trading, etc.
- Matlab version R2011a or later
- lightspeed
-
A coherent generative model for stream classification
-
The model manages its complexity by adapting the size of the latent space and the number of classifiers over time.
-
Handling concept drift by adapting data-concept association without unnecessary i.i.d. assumption among data of a batch
-
An online algorithm for inference on the non-conjugate non-parametric time-dependent models.
-
Unzip the lightspeed library and put it in the root folder of the project
-
Run the install script
-
Set the Dataset in the "Main" Script
-
Run the Main script
The results will be saved under the folder Results
In case of using the code, please cite the following paper:
Hosseini, S.A., Rabiee, H.R., Hafez, H. and Soltani-Farani, A., 2014, September. Classifying a Stream of Infinite Concepts: A Bayesian Non-parametric Approach. In Joint European Conference on Machine Learning and Knowledge Discovery in Databases (pp. 1-16). Springer Berlin Heidelberg.