Source code for PAMI.periodicFrequentPattern.basic.PFPGrowthPlus

# PFPGrowthPlus is fundamental and improved version of PFPGrowth algorithm to discover periodic-frequent patterns in temporal database.
# It uses greedy approach to discover effectively
#
# **Importing this algorithm into a python program**
# --------------------------------------------------------
#
#
#             from PAMI.periodicFrequentPattern.basic import PFPGrowthPlus as alg
#
#             obj = alg.PFPGrowthPlus("../basic/sampleTDB.txt", "2", "6")
#
#             obj.mine()
#
#             periodicFrequentPatterns = obj.getPatterns()
#
#             print("Total number of Periodic Frequent Patterns:", len(periodicFrequentPatterns))
#
#             obj.save("patterns")
#
#             Df = obj.getPatternsAsDataFrame()
#
#             memUSS = obj.getMemoryUSS()
#
#             print("Total Memory in USS:", memUSS)
#
#             memRSS = obj.getMemoryRSS()
#
#             print("Total Memory in RSS", memRSS)
#
#             run = obj.getRuntime()
#
#             print("Total ExecutionTime in seconds:", run)
#



__copyright__ = """
 Copyright (C)  2021 Rage Uday Kiran

     This program is free software: you can redistribute it and/or modify
     it under the terms of the GNU General Public License as published by
     the Free Software Foundation, either version 3 of the License, or
     (at your option) any later version.

     This program is distributed in the hope that it will be useful,
     but WITHOUT ANY WARRANTY; without even the implied warranty of
     MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
     GNU General Public License for more details.

     You should have received a copy of the GNU General Public License
     along with this program.  If not, see <https://www.gnu.org/licenses/>.
     Copyright (C)  2021 Rage Uday Kiran

"""

from PAMI.periodicFrequentPattern.basic import abstract as _ab
import pandas as pd
from deprecated import deprecated

from PAMI.periodicFrequentPattern.basic import abstract as _ab
from typing import List, Dict, Tuple, Set, Union, Any, Generator

_maxPer = float()
_minSup = float()
_lno = int()


class _Node(object):
    """
    A class used to represent the node of frequentPatternTree

    :Attributes:

        item : int
            storing item of a node
        timeStamps : list
            To maintain the timestamps of transaction at the end of the branch
        parent : node
            To maintain the parent of every node
        children : list
            To maintain the children of node

    :Methods:

        addChild(itemName)
            storing the children to their respective parent nodes
    """

    def __init__(self, item, children) -> None:
        self.item = item
        self.children = children
        self.parent = None
        self.timeStamps = []

    def addChild(self, node) -> None:
        self.children[node.item] = node
        node.parent = self


class _Tree(object):
    """
    A class used to represent the frequentPatternGrowth tree structure

    :Attributes:

        root : Node
            Represents the root node of the tree
        summaries : dictionary
            storing the nodes with same item name
        info : dictionary
            stores the support of items

    :Methods:

        addTransaction(transaction)
            creating transaction as a branch in frequentPatternTree
        getConditionalPatterns(Node)
            generates the conditional patterns from tree for specific node
        conditionalTransactions(prefixPaths,Support)
            takes the prefixPath of a node and support at child of the path and extract the frequent items from
            prefixPaths and generates prefixPaths with items which are frequent
        remove(Node)
            removes the node from tree once after generating all the patterns respective to the node
        generatePatterns(Node)
            tarts from the root node of the tree and mines the periodic-frequent patterns
    """

    def __init__(self) -> None:
        self.root = _Node(None, {})
        self.summaries = {}
        self.info = {}

    def addTransaction(self, transaction, tid) -> None:
        """
        adding transaction into tree

        :param transaction : it represents the one transaction in database
        :type transaction : list
        :param tid : represents the timestamp of transaction
        :type tid : list
        :return: None
        """
        currentNode = self.root
        for i in range(len(transaction)):
            if transaction[i] not in currentNode.children:
                newNode = _Node(transaction[i], {})
                currentNode.addChild(newNode)
                if transaction[i] in self.summaries:
                    self.summaries[transaction[i]].append(newNode)
                else:
                    self.summaries[transaction[i]] = [newNode]
                currentNode = newNode
            else:
                currentNode = currentNode.children[transaction[i]]
        currentNode.timeStamps = currentNode.timeStamps + tid

    def getConditionalPatterns(self, alpha) -> Tuple[List[List], List[List], Dict]:
        """
        generates all the conditional patterns of respective node

        :param alpha : it represents the Node in tree
        :type alpha : Node
        """
        finalPatterns = []
        finalSets = []
        for i in self.summaries[alpha]:
            set1 = i.timeStamps
            set2 = []
            while i.parent.item is not None:
                set2.append(i.parent.item)
                i = i.parent
            if len(set2) > 0:
                set2.reverse()
                finalPatterns.append(set2)
                finalSets.append(set1)
        finalPatterns, finalSets, info = self.conditionalTransactions(finalPatterns, finalSets)
        return finalPatterns, finalSets, info

    @staticmethod
    def generateTimeStamps(node) -> List:
        finalTimeStamps = node.timeStamps
        return finalTimeStamps

    def removeNode(self, nodeValue):
        """
        removing the node from tree

        :param nodeValue : it represents the node in tree
        type nodeValue : node
        """
        for i in self.summaries[nodeValue]:
            i.parent.timeStamps = i.parent.timeStamps + i.timeStamps
            del i.parent.children[nodeValue]

    def getTimeStamps(self, alpha):
        """
        to get all the timestamps related to a node in tree

        :param alpha: node of a tree
        :return: timestamps of a node
        """
        temporary = []
        for i in self.summaries[alpha]:
            temporary += i.timeStamps
        return temporary

    @staticmethod
    def getSupportAndPeriod(timeStamps):
        """
        calculates the support and periodicity with list of timestamps

        :param timeStamps : timestamps of a pattern.
        :type timeStamps : list
        """
        global _maxPer, _lno
        timeStamps.sort()
        cur = 0
        per = 0
        sup = 0
        for j in range(len(timeStamps)):
            per = max(per, timeStamps[j] - cur)
            if per > _maxPer:
                return [0, 0]
            cur = timeStamps[j]
            sup += 1
        per = max(per, _lno - cur)
        return [sup, per]

    def conditionalTransactions(self, conditionalPatterns, conditionalTimeStamps):
        """
        It generates the conditional patterns with periodic frequent items

        :param conditionalPatterns : conditionalPatterns generated from conditionalPattern method for respective node
        type conditionalPatterns : list
        :param conditionalTimeStamps : represents the timestamps of conditional patterns of a node
        :type conditionalTimeStamps : list
        """
        global _maxPer, _minSup
        pat = []
        timeStamps = []
        data1 = {}
        for i in range(len(conditionalPatterns)):
            for j in conditionalPatterns[i]:
                if j in data1:
                    data1[j] = data1[j] + conditionalTimeStamps[i]
                else:
                    data1[j] = conditionalTimeStamps[i]
        updatedDictionary = {}
        for m in data1:
            updatedDictionary[m] = self.getSupportAndPeriod(data1[m])
        updatedDictionary = {k: v for k, v in updatedDictionary.items() if v[0] >= _minSup and v[1] <= _maxPer}
        count = 0
        for p in conditionalPatterns:
            p1 = [v for v in p if v in updatedDictionary]
            trans = sorted(p1, key=lambda x: (updatedDictionary.get(x)[0], -x), reverse=True)
            if len(trans) > 0:
                pat.append(trans)
                timeStamps.append(conditionalTimeStamps[count])
            count += 1
        return pat, timeStamps, updatedDictionary

    def generatePatterns(self, prefix):
        """
        generates the patterns

        :param prefix : forms the combination of items
        :type prefix : list
        """
        for i in sorted(self.summaries, key=lambda x: (self.info.get(x)[0], -x)):
            pattern = prefix[:]
            pattern.append(i)
            yield pattern, self.info[i]
            patterns, timeStamps, info = self.getConditionalPatterns(i)
            conditionalTree = _Tree()
            conditionalTree.info = info.copy()
            for pat in range(len(patterns)):
                conditionalTree.addTransaction(patterns[pat], timeStamps[pat])
            if len(patterns) > 0:
                for q in conditionalTree.generatePatterns(pattern):
                    yield q
            self.removeNode(i)


[docs] class PFPGrowthPlus(_ab._periodicFrequentPatterns): """ :Description: PFPGrowthPlus is fundamental and improved version of PFPGrowth algorithm to discover periodic-frequent patterns in temporal database. It uses greedy approach to discover effectively :Reference: R. UdayKiran, MasaruKitsuregawa, and P. KrishnaReddyd, "Efficient discovery of periodic-frequent patterns in very large databases," Journal of Systems and Software February 2016 https://doi.org/10.1016/j.jss.2015.10.035 :param iFile: str : Name of the Input file to mine complete set of periodic frequent pattern's :param oFile: str : Name of the output file to store complete set of periodic frequent pattern's :param minSup: str: Controls the minimum number of transactions in which every item must appear in a database. :param maxPer: str: Controls the maximum number of transactions in which any two items within a pattern can reappear. :param sep: str : This variable is used to distinguish items from one another in a transaction. The default seperator is tab space. However, the users can override their default separator. :Attributes: iFile : file Name of the Input file or path of the input file oFile : file Name of the output file or path of the output file minSup : int or float or str The user can specify minSup either in count or proportion of database size. If the program detects the data type of minSup is integer, then it treats minSup is expressed in count. Otherwise, it will be treated as float. Example: minSup=10 will be treated as integer, while minSup=10.0 will be treated as float maxPer : int or float or str The user can specify maxPer either in count or proportion of database size. If the program detects the data type of maxPer is integer, then it treats maxPer is expressed in count. Otherwise, it will be treated as float. Example: maxPer=10 will be treated as integer, while maxPer=10.0 will be treated as float sep : str This variable is used to distinguish items from one another in a transaction. The default seperator is tab space or \t. However, the users can override their default separator. memoryUSS : float To store the total amount of USS memory consumed by the program memoryRSS : float To store the total amount of RSS memory consumed by the program startTime:float To record the start time of the mining process endTime:float To record the completion time of the mining process Database : list To store the transactions of a database in list mapSupport : Dictionary To maintain the information of item and their frequency lno : int it represents the total no of transaction tree : class it represents the Tree class itemSetCount : int it represents the total no of patterns finalPatterns : dict it represents to store the patterns :Methods: mine() Mining process will start from here getPatterns() Complete set of patterns will be retrieved with this function save(oFile) Complete set of periodic-frequent patterns will be loaded in to a output file getPatternsAsDataFrame() Complete set of periodic-frequent patterns will be loaded in to a dataframe getMemoryUSS() Total amount of USS memory consumed by the mining process will be retrieved from this function getMemoryRSS() Total amount of RSS memory consumed by the mining process will be retrieved from this function getRuntime() Total amount of runtime taken by the mining process will be retrieved from this function check(line) To check the delimiter used in the user input file creatingItemSets(fileName) Scans the dataset or dataframes and stores in list format PeriodicFrequentOneItem() Extracts the one-periodic-frequent patterns from Databases updateDatabases() update the Databases by removing aperiodic items and sort the Database by item decreased support buildTree() after updating the Databases ar added into the tree by setting root node as null mine() the main method to run the program **Methods to execute code on terminal** ------------------------------------------- .. code-block:: console Format: (.venv) $ python3 PFPGrowthPlus.py <inputFile> <outputFile> <minSup> <maxPer> Example: (.venv) $ python3 PFPGrowthPlus.py sampleTDB.txt patterns.txt 0.3 0.4 .. note:: minSup will be considered in percentage of database transactions **Importing this algorithm into a python program** ----------------------------------------------------------- .. code-block:: python from PAMI.periodicFrequentPattern.basic import PFPGorwthPlus as alg obj = alg.PFPGrowthPlus("../basic/sampleTDB.txt", "2", "6") obj.mine() periodicFrequentPatterns = obj.getPatterns() print("Total number of Periodic Frequent Patterns:", len(periodicFrequentPatterns)) obj.save("patterns") Df = obj.getPatternsAsDataFrame() memUSS = obj.getMemoryUSS() print("Total Memory in USS:", memUSS) memRSS = obj.getMemoryRSS() print("Total Memory in RSS", memRSS) run = obj.getRuntime() print("Total ExecutionTime in seconds:", run) **Credits:** ----------------- The complete program was written by P.Likhitha under the supervision of Professor Rage Uday Kiran. """ _minSup = str() _maxPer = str() _startTime = float() _endTime = float() _finalPatterns = {} _iFile = " " _oFile = " " _sep = " " _memoryUSS = float() _memoryRSS = float() _Database = [] _rank = {} _rankedUp = {} _lno = 0 def _creatingItemSets(self) -> None: """ Storing the complete transactions of the database/input file in a database variable :return: None """ self._Database = [] if isinstance(self._iFile, _ab._pd.DataFrame): data, ts = [], [] if self._iFile.empty: print("its empty..") i = self._iFile.columns.values.tolist() if 'TS' in i: ts = self._iFile['TS'].tolist() if 'Transactions' in i: data = self._iFile['Transactions'].tolist() for i in range(len(data)): if data[i]: tr = [str(ts[i])] + [x for x in data[i].split(self._sep)] self._Database.append(tr) else: self._Database.append([str(ts[i])]) if isinstance(self._iFile, str): if _ab._validators.url(self._iFile): data = _ab._urlopen(self._iFile) for line in data: line.strip() line = line.decode("utf-8") temp = [i.rstrip() for i in line.split(self._sep)] temp = [x for x in temp if x] self._Database.append(temp) else: try: with open(self._iFile, 'r', encoding='utf-8') as f: for line in f: line.strip() temp = [i.rstrip() for i in line.split(self._sep)] temp = [x for x in temp if x] self._Database.append(temp) except IOError: print("File Not Found") quit() maxNos = [int(x[0]) for x in self._Database] self._lno = max(maxNos) def _periodicFrequentOneItem(self) -> Tuple[Dict, List]: """ calculates the support of each item in the dataset and assign the ranks to the items by decreasing support and returns the frequent items list """ data = {} for tr in self._Database: n = int(tr[0]) for i in range(1, len(tr)): if n <= self._maxPer: if tr[i] not in data: data[tr[i]] = [int(tr[0]), int(tr[0]), 1] else: data[tr[i]][0] = max(data[tr[i]][0], (int(tr[0]) - data[tr[i]][1])) data[tr[i]][1] = int(tr[0]) data[tr[i]][2] += 1 else: if tr[i] in data: lp = abs(n - data[tr[i]][1]) if lp > self._maxPer: del data[tr[i]] else: data[tr[i]][0] = max(data[tr[i]][0], lp) data[tr[i]][1] = int(tr[0]) data[tr[i]][2] += 1 for key in data: data[key][0] = max(data[key][0], _lno - data[key][1]) data = {k: [v[2], v[0]] for k, v in data.items() if v[0] <= self._maxPer and v[2] >= self._minSup} genList = [k for k, v in sorted(data.items(), key=lambda x: (x[1][0], x[0]), reverse=True)] self._rank = dict([(index, item) for (item, index) in enumerate(genList)]) # genList=[k for k,v in sorted(data.items(),key=lambda x: (x[1][0],x[0]),reverse=True)] return data, genList def _updateTransactions(self, dict1) -> List: """ remove the items which are not frequent from transactions and updates the transactions with rank of items :param dict1 : frequent items with support :type dict1 : dictionary """ list1 = [] for tr in self._Database: list2 = [int(tr[0])] for i in range(1, len(tr)): if tr[i] in dict1: list2.append(self._rank[tr[i]]) if len(list2) >= 2: basket = list2[1:] basket.sort() list2[1:] = basket[0:] list1.append(list2) return list1 @staticmethod def _buildTree(data, info) -> _Tree: """ It takes the transactions and support of each item and construct the main tree with setting root node as null :param data : it represents the one transaction in database :type data : list :param info : it represents the support of each item :type info : dictionary """ rootNode = _Tree() rootNode.info = info.copy() for i in range(len(data)): set1 = [data[i][0]] rootNode.addTransaction(data[i][1:], set1) return rootNode def _savePeriodic(self, itemSet) -> str: """ To convert item ranks into original item names :param itemSet: periodic-frequent pattern :return: original itemSet """ t1 = str() for i in itemSet: t1 = t1 + self._rankedUp[i] + "\t" return t1 def _convert(self, value) -> Union[int, float]: """ To convert the given user specified value :param value: user specified value :return: converted value """ if type(value) is int: value = int(value) if type(value) is float: value = (len(self._Database) * value) if type(value) is str: if '.' in value: value = float(value) value = (len(self._Database) * value) else: value = int(value) return value
[docs] def mine(self) -> None: """ Main method where the patterns are mined by constructing tree. :return: None """ global _minSup, _maxPer, _lno self._startTime = _ab._time.time() if self._iFile is None: raise Exception("Please enter the file path or file name:") if self._minSup is None: raise Exception("Please enter the Minimum Support") self._creatingItemSets() self._minSup = self._convert(self._minSup) self._maxPer = self._convert(self._maxPer) _minSup, _maxPer, _lno = self._minSup, self._maxPer, len(self._Database) generatedItems, pfList = self._periodicFrequentOneItem() updatedTransactions = self._updateTransactions(generatedItems) for x, y in self._rank.items(): self._rankedUp[y] = x info = {self._rank[k]: v for k, v in generatedItems.items()} Tree = self._buildTree(updatedTransactions, info) patterns = Tree.generatePatterns([]) self._finalPatterns = {} for i in patterns: x = self._savePeriodic(i[0]) self._finalPatterns[x] = i[1] self._endTime = _ab._time.time() process = _ab._psutil.Process(_ab._os.getpid()) self._memoryRSS = float() self._memoryUSS = float() self._memoryUSS = process.memory_full_info().uss self._memoryRSS = process.memory_info().rss print("periodic-frequent patterns were generated successfully using PFPGrowth++ algorithm ")
[docs] def startMine(self) -> None: """ Main method where the patterns are mined by constructing tree. :return: None """ self.mine()
# global _minSup, _maxPer, _lno # self._startTime = _ab._time.time() # if self._iFile is None: # raise Exception("Please enter the file path or file name:") # if self._minSup is None: # raise Exception("Please enter the Minimum Support") # self._creatingItemSets() # self._minSup = self._convert(self._minSup) # self._maxPer = self._convert(self._maxPer) # _minSup, _maxPer, _lno = self._minSup, self._maxPer, len(self._Database) # generatedItems, pfList = self._periodicFrequentOneItem() # updatedTransactions = self._updateTransactions(generatedItems) # for x, y in self._rank.items(): # self._rankedUp[y] = x # info = {self._rank[k]: v for k, v in generatedItems.items()} # Tree = self._buildTree(updatedTransactions, info) # patterns = Tree.generatePatterns([]) # self._finalPatterns = {} # for i in patterns: # x = self._savePeriodic(i[0]) # self._finalPatterns[x] = i[1] # self._endTime = _ab._time.time() # process = _ab._psutil.Process(_ab._os.getpid()) # self._memoryRSS = float() # self._memoryUSS = float() # self._memoryUSS = process.memory_full_info().uss # self._memoryRSS = process.memory_info().rss # print("periodic-frequent patterns were generated successfully using PFPGrowth++ algorithm ")
[docs] def getMemoryUSS(self) -> float: """ Total amount of USS memory consumed by the mining process will be retrieved from this function :return: returning USS memory consumed by the mining process :rtype: float """ return self._memoryUSS
[docs] def getMemoryRSS(self) -> float: """ Total amount of RSS memory consumed by the mining process will be retrieved from this function :return: returning RSS memory consumed by the mining process :rtype: float """ return self._memoryRSS
[docs] def getRuntime(self) -> float: """ Calculating the total amount of runtime taken by the mining process :return: returning total amount of runtime taken by the mining process :rtype: float """ return self._endTime - self._startTime
[docs] def getPatternsAsDataFrame(self) -> _ab._pd.DataFrame: """ Storing final periodic-frequent patterns in a dataframe :return: returning periodic-frequent patterns in a dataframe :rtype: pd.DataFrame """ dataframe = {} data = [] for a, b in self._finalPatterns.items(): data.append([a, b[0], b[1]]) dataframe = _ab._pd.DataFrame(data, columns=['Patterns', 'Support', 'Periodicity']) return dataframe
[docs] def save(self, outFile: str) -> None: """ Complete set of periodic-frequent patterns will be loaded in to an output file :param outFile: name of the output file :type outFile: csv file :return: None """ self._oFile = outFile writer = open(self._oFile, 'w+') for x, y in self._finalPatterns.items(): s1 = x + ":" + str(y[0]) + ":" + str(y[1]) #s1 = x.replace(' ','\t').strip() + ":" + str(y[0]) + ":" + str(y[1]) writer.write("%s \n" % s1)
[docs] def getPatterns(self) -> Dict[str, Tuple[int, int]]: """ Function to send the set of periodic-frequent patterns after completion of the mining process :return: returning periodic-frequent patterns :rtype: dict """ return self._finalPatterns
[docs] def printResults(self) -> None: """ This function is used to print the results :return: None """ print("Total number of Periodic Frequent Patterns:", len(self.getPatterns())) print("Total Memory in USS:", self.getMemoryUSS()) print("Total Memory in RSS", self.getMemoryRSS()) print("Total ExecutionTime in ms:", self.getRuntime())
if __name__ == "__main__": _ap = str() if len(_ab._sys.argv) == 5 or len(_ab._sys.argv) == 6: if len(_ab._sys.argv) == 6: _ap = PFPGrowthPlus(_ab._sys.argv[1], _ab._sys.argv[3], _ab._sys.argv[4], _ab._sys.argv[5]) if len(_ab._sys.argv) == 5: _ap = PFPGrowthPlus(_ab._sys.argv[1], _ab._sys.argv[3], _ab._sys.argv[4]) _ap.mine() print("Total number of Periodic-Frequent Patterns:", len(_ap.getPatterns())) _ap.save(_ab._sys.argv[2]) print("Total Memory in USS:", _ap.getMemoryUSS()) print("Total Memory in RSS", _ap.getMemoryRSS()) print("Total ExecutionTime in ms:", _ap.getRuntime()) else: print("Error! The number of input parameters do not match the total number of parameters provided")