Pattern recognition


1. Introduction. Recognition of patterns.  
2. Measures of prediction  
3. The H-method of estimation.  
4. Distance measures.  
5. Graphic analysis.  
6. Case study  
7. Hierarchical clustering.  
8. K-means clustering.  
9. Examples  
10. Dimension analysis.  
11. Error rate analysis.  
12. Distance measures.  
13. Importance of variables.  
14. Quadratic procedures  
15. Trends in data.  
16. Regression trees  
17. Confidence intervals  
18. Gravity centers  
19. Sensitivity analysis  
20. Case study  
21. Guidelines for presentation of results  

Pattern recognition is important in many situations, both in applied sciences and industry. The approach of the H-method can be extended to the area of pattern recognition. The basic idea of the H-method is to generate weights of the variables such that the resulting 'profile vector' (score vector) has predictive performance. The results at each step is thus a 'representative' that is able to describe the data to some degree. Each profile vector can be divided into parts and several ones can be generated to represent the parts of data. This flexibility of generating one or more profile vectors from subsets of data makes it possible to recognize different patterns in data. This approach can be combined with other strategies of analysis.

An example is regression trees.