10-23-2019, 07:43 PM
Topic- A comparative analysis of various clustering algorithms in data mining
University - RTMNU(rashtrasant tukadoji maharaj nagpur university)
Project guide lines
Form the 5 types clustering algorithm partitioning based, hierarchical based, density based , grid based, model based we have to take one method from each of the types,
For ex. 1. Partitioning based :- k-means
2. Hierarchical based :- BIRCH
3. Density based :- DBSCAN
4. Grid based :- CLIQUE or STING or Wave cluster
5. Model based :- EM or COMWEB or SLINK
Then the chosen algorithms has to be ompared on the following parameters:
Time complexity
Space complexity
Size of clusters
Number of clusters
Shape of clusters
Density threshold
Which data set is suitable
Non-convex clusters
Agglomerative
Accuracy
The steps will be like after running the code a gui box will appear where there it will ask which algorithm/algorithms to compare the below that a checkbox to compare of what parameters will be asked then at last with figures and table and some details the comparison should be displayed.
University - RTMNU(rashtrasant tukadoji maharaj nagpur university)
Project guide lines
Form the 5 types clustering algorithm partitioning based, hierarchical based, density based , grid based, model based we have to take one method from each of the types,
For ex. 1. Partitioning based :- k-means
2. Hierarchical based :- BIRCH
3. Density based :- DBSCAN
4. Grid based :- CLIQUE or STING or Wave cluster
5. Model based :- EM or COMWEB or SLINK
Then the chosen algorithms has to be ompared on the following parameters:
Time complexity
Space complexity
Size of clusters
Number of clusters
Shape of clusters
Density threshold
Which data set is suitable
Non-convex clusters
Agglomerative
Accuracy
The steps will be like after running the code a gui box will appear where there it will ask which algorithm/algorithms to compare the below that a checkbox to compare of what parameters will be asked then at last with figures and table and some details the comparison should be displayed.