This section introduces several basic methods in pattern recognition (PR for short), which tries to construct a classifier or recognizer that can classify unseen data based on a given dataset with known classes. In general, we have two datasets for PR:
The general approach to PR involves the following steps:
- Training set: This is the dataset with known classes. We use this dataset to construct the best classifier via different methods.
- Test set: This is the unseen dataset not used for constructing the classifier. We need this dataset to verify the performance of the constructed classifier.
- Data collection and feature selection
- Divide the data into training and test sets
- Select a method to construct the classifier based on the training set
- Verify the performance of the method based on the test set
- If the performance is satisfactory, stop. Otherwise, go back to step 3 or step 1.
Data Clustering and Pattern Recognition (資料分群與樣式辨認)