Python Wrapper for SPMF Java library.
This module contains python wrappers for pattern mining algorithms implemented in SPMF Java library. Each algorithm is implemented as a standalone Python class with fully descriptive and tested APIs. It also provides native support for Pandas dataframes.
Why? If you're in a Python pipeline, it might be cumbersome to use Java as an intermediate step. Using spmf-wrapper you can stay in your pipeline as though Java is never used at all.
A Java Runtime Environment is required to run this wrapper. If an existing installation is not detected, JRE v21 is automatically installed using install-jdk python module at $HOME/.jre/jdk-21.0.2+13-jre. If you prefer to install Java Runtime manually, follow instructions here. Test installation by running the following command on the terminal:
> java -version
java version "1.8.0_391"
Java(TM) SE Runtime Environment (build 1.8.0_391-b13)
Java HotSpot(TM) 64-Bit Server VM (build 25.391-b13, mixed mode)
Example:
from spmf import EMMA
emma = EMMA(min_support=2, max_window=2, timestamp_present=True, transform=True)
output = emma.run_pandas(input_df)Input:
| Time points | Itemset | |
|---|---|---|
| 0 | 1 | a |
| 1 | 2 | a |
| 2 | 3 | a |
| 3 | 3 | b |
| 4 | 6 | a |
| 5 | 7 | a |
| 6 | 7 | b |
| 7 | 8 | c |
| 8 | 9 | b |
| 9 | 11 | d |
Output:
| Frequent episode | Support | |
|---|---|---|
| 0 | a | 5 |
| 1 | b | 3 |
| 2 | a b | 2 |
| 3 | a-> a | 3 |
| 4 | a -> b | 2 |
| 5 | a -> a b | 2 |
See examples for more details.
For a detailed explanation of the algorithm and parameters, refer to the corresponding webpage in the SPMF documentation.
| Algorithm | Type | Implemented |
|---|---|---|
| PrefixSpan | Frequent Sequential Pattern | ✓ |
| GSP | Frequent Sequential Pattern | |
| SPADE | Frequent Sequential Pattern | ✓ |
| CM-SPADE | Frequent Sequential Pattern | ✓ |
| SPAM | Frequent Sequential Pattern | ✓ |
| CM-SPAM | Frequent Sequential Pattern | |
| FAST | Frequent Sequential Pattern | |
| LAPIN | Frequent Sequential Pattern | |
| ClaSP | Frequent Closed Sequential Pattern | ✓ |
| CM-ClaSP | Frequent Closed Sequential Pattern | ✓ |
| CloFAST | Frequent Closed Sequential Pattern | |
| CloSpan | Frequent Closed Sequential Pattern | |
| BIDE+ | Frequent Closed Sequential Pattern | |
| Post Processing SPAM or PrefixSpan | Frequent Closed Sequential Pattern | |
| MaxSP | Frequent Maximal Sequential Pattern | |
| VMSP | Frequent Maximal Sequential Pattern | ✓ |
| FEAT | Frequent Sequential Generator Pattern | |
| FSGP | Frequent Sequential Generator Pattern | |
| VGEN | Frequent Sequential Generator Pattern | ✓ |
| NOSEP | Non-overlapping Sequential Pattern | ✓ |
| GoKrimp | Compressing Sequential Pattern | |
| TKS | Top-k Frequent Sequential Pattern | ✓ |
| TSP | Top-k Frequent Sequential Pattern |
| Algorithm | Type | Implemented |
|---|---|---|
| EMMA | Frequent Episode | ✓ |
| AFEM | Frequent Episode | ✓ |
| MINEPI | Frequent Episode | |
| MINEPI+ | Frequent Episode | ✓ |
| TKE | Top-k Frequent Episodes | ✓ |
| MaxFEM | Maximal Frequent Episodes | ✓ |
| POERM | Episode Rules | |
| POERM-ALL | Episode Rules | |
| POERMH | Episode Rules | |
| NONEPI | Episode Rules | ✓ |
| TKE-Rules | Episode Rules | ✓ |
| AFEM-Rules | Episode Rules | ✓ |
| EMMA-Rules | Epsiode Rules | ✓ |
| MINEPI+-Rules | Episode Rules | |
| HUE-SPAN | High Utility Episodes | |
| US-SPAN | High Utility Episodes | |
| TUP | Top-K High Utility Episodes |
Fournier-Viger, P., Lin, C.W., Gomariz, A., Gueniche, T., Soltani, A., Deng, Z., Lam, H. T. (2016).
The SPMF Open-Source Data Mining Library Version 2.
Proc. 19th European Conference on Principles of Data Mining and Knowledge Discovery (PKDD 2016) Part III, Springer LNCS 9853, pp. 36-40.
