IM Theses, 2001
http://www.cs.ait.ac.th/xmlui/handle/123456789/77
2020-05-25T12:18:14ZA Further Study on Backpropagation and B-Spline Neural Networks in Filtering and Forecasting Problems
http://www.cs.ait.ac.th/xmlui/handle/123456789/759
A Further Study on Backpropagation and B-Spline Neural Networks in Filtering and Forecasting Problems
Kha, Nguyen Duc Anh
2001-01-01T00:00:00ZA Comparison of Selected Training Algorithms for Recurrent Neural Networks
http://www.cs.ait.ac.th/xmlui/handle/123456789/758
A Comparison of Selected Training Algorithms for Recurrent Neural Networks
Chairatanatrai, Aree
2001-12-01T00:00:00ZMachine learning for market segmentation
http://www.cs.ait.ac.th/xmlui/handle/123456789/527
Machine learning for market segmentation
Rujiphorn Techathaweerit
Market Segmentation is the process of identifying groups of customers that have similarities in characteristics or similarities in needs. Segmenting the market can help firms increase profits by better targeting advertising of products and services. Customer preferences are one of the most attractive bases for segmentation. The main problem is to apply appropriate algorithms to segment customers based on preferences.
In this thesis, techniques from machine learning and collaborative filtering are integrated to develop a comprehensive methodology for segmenting customers based on their preferences. The effectiveness of the approach is evaluated on a database of movie preferences. A particular similarity measure technique from collaborative filtering is selected for calculating the similarity between users. The clustering algorithms are responsible for clustering the users based on the similarity. Finally, Bayesian multi nets and decision trees are used for generating predictive models of the segments. In addition, decision trees are used for creating descriptions of the segments.
87 p.
2001-01-01T00:00:00ZRelationships between BERNSTEIN, CHEBYSHEV and LEGENDRE models
http://www.cs.ait.ac.th/xmlui/handle/123456789/526
Relationships between BERNSTEIN, CHEBYSHEV and LEGENDRE models
Sathasivam Amirthalingam
This study was carried out with the main objective of establishing the relationships between Bezier, Chebyshev and Legendre curves. These relationships are obtained using the polar form approach with mathematical derivations. They can serve as convenient ways to convert a curve in a given form into the other two remaining forms.
It was found that
- the control points of a curve given in one of the forms Bezier, Chebyshev and Legendre, can be expressed as affine combinations of the control points of in each of the remaining forms for the rational case. This result is not true for non-rational curves.
- the control points of a Chebyshev or Legendre curve do not provide any hint about the shape of the curve defined by them. This is completely different from the case of Bezier curves.
In view of these results, it seems that when a curve is given in the Chebyshev’s or Legendre’s form, it should be converted into the Bezier form so that its shape can be roughly figured out before any drawing is carried out.
127 p.
2001-01-01T00:00:00Z