clustering similar solutions
to run clustering for the desired puzzle:
cd puzzlesdata/SRC
python3 strategyClastur.pyEnter the puzzle number:
Enter the number of clusters:
Enter the path of the folder containing pnp the json files:
The results is saved in /puzzlesdata/Plots_Text/clustering/puzzle {puzzleNumber}
note:
- clustering methode is based on scipy.cluster.hierarchy.linkage and "ward" methode for cluster distances
- distance between each solution is calculated by dtw
- there are parameters in each of the methos as well as number of clusters that need to be determined
- the labels of interaction types are transformed by oneHotEncoding so that each interaction type has same distance to another one
earlier results
as requsted here:
The number of participants who solved all the puzzles: 12 out of 15
The participants who eventually solved all the puzzles: [31 33 34 36 39 41 42 43 45 46 47 48]
The participants who did not solve all the puzzles: [35 37 44]
The participant 35 solved 26 puzzles out of 27
The puzzles that the participant 35 did not solve are: [27]
The participant 37 solved 25 puzzles out of 27
The puzzles that the participant 37 did not solve are: [10 27]
The participant 44 solved 20 puzzles out of 27
The puzzles that the participant 44 did not solve are: [ 7 8 11 21 24 25 27]
With stats , it is feasible now to request any form of general stats of the data needed