FWPODS PY

A python implementation library based on the paper named "A sliding window-based approach for mining frequent weighted patterns over data streams

A

H. Bui, T. -A. Nguyen-Hoang, B. Vo, H. Nguyen and T. Le, "A Sliding Window-Based Approach for Mining Frequent Weighted Patterns Over Data Streams," in IEEE Access, vol. 9, pp. 56318-56329, 2021, doi: 10.1109/ACCESS.2021.3070132. keywords: {Data mining;Data models;Databases;Urban areas;Itemsets;Mathematical model;Information technology;Pattern mining;data streams;frequent weighted patterns;sliding window model},

Example usage

Using the existing window manager to add a transaction to the window and start the mining process

from random import randint

from collections import OrderedDict

from fwpods_py.classes import *



FWPs = []

runtimes = []

item_weights = {}



window_size = 45000

min_ws = 0.8

panel_size = 1

twm = weights_manager()



transactions = OrderedDict()

count = 0

# This sample dataset can be found on the SPMF website at https://www.philippe-fournier-viger.com/spmf/index.php?link=datasets.php

ds_name = "retail"

with open(f"./datasets/{ds_name}.txt", "r") as f:

    t_id = "1"

    for line in f:

        if count == window_size + 50:

            break

        transactions[t_id] = line.strip().split()

        for item in line.strip().split():

            if item in item_weights:

                continue

            else:

                item_weights[item] = randint(1, 10)

        t_id = str(int(t_id) + 1)

        count += 1



win_man = window_manager(None, window_size, panel_size, min_ws)

win_man.new_weights(item_weights)



# Simulate a data stream

for t_id, t_items in transactions.items():

    win_man.add_transaction(t_id, t_items)



res_location = f"./results/{ds_name}/"

with open(f"{ds_name}_runtime_total.txt", "w") as f:

    for ttr in win_man.total_runtime:

        f.write(f"{ttr.total_seconds()}\n")



with open(f"{ds_name}_runtime_algo.txt", "w") as f:

    for art in win_man.algo_runtime:

        f.write(f"{art.total_seconds()}\n")



with open(f"{ds_name}_runtime_tree.txt", "w") as f:

    for tr in win_man.tree_build_time:

        f.write(f"{tr.total_seconds()}\n")