技術標籤：java演算法FPGrowth資料探勘關聯規則

實驗結果

最小支援度0.001條件下可以得到準確結果，僅用1.6s
最小可以支援最小支援度為0.0003的計算

程式碼

package com.company;

import java.io.*;
import java.util.*;

class FPNode {
    String name;
    int count = 0;
    FPNode brother = null, parent = null;
    ArrayList<FPNode> children = new ArrayList<>
 
();

    FPNode(String name) {
        this.name = name;
    }

    FPNode(String name, FPNode parent, int count) {
        this.name = name;
        this.parent = parent;
        this.count = count;
    }
}

class Item {
    String name;
    int support;

    Item(String name, int support) {
        this
 
.name = name;
        this.support = support;
    }
}

class Trade {
    ArrayList<Item> items = new ArrayList<>();

    public Trade(List<String> items, Map<String, Integer> frequentItems) {
        items.forEach(item -> {
            if (frequentItems.containsKey(item))
                this
 
.items.add(new Item(item, frequentItems.get(item)));
        });
        this.items.sort((l, r) -> Integer.compare(r.support, l.support));//按支援度從大到小排序
    }
}

class FPTree {
    int minSupport;
    FPNode root = new FPNode("root");
    Map<String, Integer> frequentItems = new HashMap<>();
    Map<String, FPNode> headTable = new HashMap<>();
    //該項對應的最後一個節點
    Map<String, FPNode> currentPosition = new HashMap<>();
    List<List<String>> database;

    public FPTree(List<List<String>> database, int minSupport) {
        this.minSupport = minSupport;
        this.database = database;
        getFrequentItems();
        buildHeadTable();
        buildTree();
    }

    //計算支援度
    public void getFrequentItems() {
        Map<String, Integer> supportCount = new HashMap<>();
        for (List<String> line :
                database)
            for (String item :
                    line)
                supportCount.merge(item, 1, Integer::sum);
        supportCount.forEach((item, support) -> {
            if (support >= minSupport)
                frequentItems.put(item, support);
        });
    }

    //建立頭表
    public void buildHeadTable() {
        frequentItems.keySet().forEach(frequentItem -> {
            headTable.put(frequentItem, new FPNode(frequentItem));
            currentPosition.put(frequentItem, headTable.get(frequentItem));
        });
    }

    //建立FP-tree
    public void buildTree() {
        database.forEach(items -> {
            Trade trade = new Trade(items, frequentItems);
            insertTree(trade, root);
        });
    }

    public int insertTree(Trade trade, FPNode fpNode) {
        if (trade.items.size() == 0)
            return 0;
        Item item = trade.items.remove(0);
        FPNode nextFPNode;
        // 查詢該項是否存在
        for (FPNode children :
                fpNode.children)
            if (children.name.equals(item.name)) {
                children.count++;
                nextFPNode = children;
                return insertTree(trade, nextFPNode);
            }
        nextFPNode = new FPNode(item.name, fpNode, 1);
        fpNode.children.add(nextFPNode);
        currentPosition.get(item.name).brother = nextFPNode;
        currentPosition.put(item.name, nextFPNode);
        return insertTree(trade, nextFPNode);
    }
}

class FrequentItem {
    List<String> items;
    int support;

    public FrequentItem(List<String> items, int support) {
        this.items = items;
        this.support = support;
    }
}

class FPGrowth {
    int minSupport;
    List<FrequentItem> patternList = new ArrayList<>();

    public FPGrowth(FPTree fpTree, int minSupport) {
        this.minSupport = minSupport;
        fpGrowth(fpTree, null);
    }

    private void fpGrowth(FPTree fpTree, List<String> suffix) {
        if (fpTree.root.children.size() == 0)
            return;
        fpTree.frequentItems.keySet().forEach(frequentItem -> {
            List<String> newSuffix = new ArrayList<>();
            newSuffix.add(frequentItem);
            if (suffix != null && !suffix.isEmpty())
                newSuffix.addAll(suffix);
            patternList.add(new FrequentItem(newSuffix, fpTree.frequentItems.get(frequentItem)));
            //生成條件模式庫
            List<List<String>> conditionalPatternDatabase = generateConditionalPatternDatabase(fpTree, frequentItem);
            //生成條件FPTree
            FPTree conditionalFPTree = new FPTree(conditionalPatternDatabase, minSupport);
            fpGrowth(conditionalFPTree, newSuffix);
        });
    }

    private List<List<String>> generateConditionalPatternDatabase(FPTree fpTree, String frequentItem) {
        List<List<String>> conditionalPatternDatabase = new ArrayList<>();
        FPNode headNode = fpTree.headTable.get(frequentItem);
        for (FPNode fpNode = headNode.brother; fpNode != null; fpNode = fpNode.brother) {
            //生成字首路徑
            List<String> prefixPath = new ArrayList<>();
            for (FPNode fpNode1 = fpNode.parent; fpNode1.parent != null; fpNode1 = fpNode1.parent)
                prefixPath.add(fpNode1.name);
            for (int i = 0; i < fpNode.count; i++)
                conditionalPatternDatabase.add(prefixPath);
        }
        return conditionalPatternDatabase;
    }

    public List<FrequentItem> getPatternList() {
        patternList.sort((l, r) -> Integer.compare(r.support, l.support));
        return patternList;
    }
}

public class Main {
    private static final List<List<String>> database = new ArrayList<>();

    public static void main(String[] args) throws IOException {
        double minSupport;
        int count = 0;
        Scanner scanner = new Scanner(System.in);
        minSupport = scanner.nextDouble();
        long startTime = System.currentTimeMillis();
        loadData();
        FPTree fpTree = new FPTree(database, (int) Math.ceil(minSupport * database.size()));
        FPGrowth fpGrowth = new FPGrowth(fpTree, (int) Math.ceil(minSupport * database.size()));
        for (FrequentItem frequentItem :
                fpGrowth.getPatternList()) {
            System.out.println(frequentItem.items + ": " + frequentItem.support);
            count++;
        }
        System.out.println("總數: " + count);
        long endTime = System.currentTimeMillis();
        System.out.println("程式執行時間：" + (endTime - startTime) + "ms");
    }

    private static void loadData() throws IOException {
        try (BufferedReader bufferedReader = new BufferedReader(new FileReader("retail.dat"))) {
            String line;
            while ((line = bufferedReader.readLine()) != null) {
                String[] temp = line.split(" ");
                database.add(Arrays.asList(temp));
            }
        }
    }
}