balancer8m
03-24-2011, 06:15 AM
you can own to implement a SortedList, support for duplicate keys , of course, that , like Microsoft should do a
Thank Pharaoh and Ninputer
known problems and solutions :
questions : dbscan.cs line 64,power balance wristband (http://www.powerbalancebracelet.ca), SortedList does not support duplicate keys , so if two data points from the same can not be added to the collection
solution: artificially reduced by a Small amount of the data points and do not affect the clustering results from different
problem on a solution : reduce the small amount of help to make double.Epsilon SortedList that the distance between two points and the different
solution: use a Small quantity of the exponential growth , continuous retry until the SortedList that the distance has different
recommendations , both methods can work, but I finally adopted a Pharaoh 's technology programs, technology is not the best, only better thing , and now this implementation and to resolve all of the following problems, Reserved only for nostalgic below ,power balance bracelet (http://www.powerbalancebandsuk.com), the code has been updated .
DBSCAN implementation code and sample downloads : dbscan
Examination because of the previous stage , did not write long blog , and I joined the business intelligence team has never been out after a decent blog, I hope this CSharp on DBSCAN clustering algorithm to achieve the blog can cause you some concern.
description of the procedures :
Form1.cs is the application of clustering algorithm DBSCAN (Density-Based Spatical Clustering of Application with Noise) of the sample , two parameters can be adjusted EPS and MinPts clustering. DBSCAN.cs is all the algorithm implementation document clustering algorithm for further information,balance bands (http://www.powerbalancebandsuk.com), please refer to Cluster sample data from the sxdb.mdb, an Access database.
suggestions for further improvement :
may double the force through transformation of byte type of memory (assuming double type into 8 byte)
and then subtracting the last byte 0x01 to solve the problem can be more beautiful , but ... ... Oh , C # , I will not this operation
Thank Pharaoh and Ninputer
known problems and solutions :
questions : dbscan.cs line 64,power balance wristband (http://www.powerbalancebracelet.ca), SortedList does not support duplicate keys , so if two data points from the same can not be added to the collection
solution: artificially reduced by a Small amount of the data points and do not affect the clustering results from different
problem on a solution : reduce the small amount of help to make double.Epsilon SortedList that the distance between two points and the different
solution: use a Small quantity of the exponential growth , continuous retry until the SortedList that the distance has different
recommendations , both methods can work, but I finally adopted a Pharaoh 's technology programs, technology is not the best, only better thing , and now this implementation and to resolve all of the following problems, Reserved only for nostalgic below ,power balance bracelet (http://www.powerbalancebandsuk.com), the code has been updated .
DBSCAN implementation code and sample downloads : dbscan
Examination because of the previous stage , did not write long blog , and I joined the business intelligence team has never been out after a decent blog, I hope this CSharp on DBSCAN clustering algorithm to achieve the blog can cause you some concern.
description of the procedures :
Form1.cs is the application of clustering algorithm DBSCAN (Density-Based Spatical Clustering of Application with Noise) of the sample , two parameters can be adjusted EPS and MinPts clustering. DBSCAN.cs is all the algorithm implementation document clustering algorithm for further information,balance bands (http://www.powerbalancebandsuk.com), please refer to Cluster sample data from the sxdb.mdb, an Access database.
suggestions for further improvement :
may double the force through transformation of byte type of memory (assuming double type into 8 byte)
and then subtracting the last byte 0x01 to solve the problem can be more beautiful , but ... ... Oh , C # , I will not this operation