I used a total of 13 datasets(from UC Irvine repository) to validate the algorithm and visualized them using pyplot.
Key observation: As the minimum support value increases the time taken by the algo decreases. It implies that, with a lower support the fp-algo must run the mining for the bigger conditional datasets.
Steps included in this algorithm:
- Construct FP tree
- Insert FP tree
- Mine FP tree recursively
- Conditional pattern base
- Creating frequent pattern item sets
Implementation:
- Data preprocessing
- Creating and updating FP tree
- Mining and creating conditional FP tree