Apriori算法

        刚开始接触Java，再加上学了《数据挖掘》，就用Java实现了Apriori关联规则算法，花了我今天一天的时间，眼睛都看酸了。

import java.text.DecimalFormat;
import java.util.ArrayList;
import java.util.Collections;

public class Apriori {
	
	//产生候选集
	public ArrayList<Item> CreateCandidateSet(ArrayList<Item> FrequentItemSet, String DataSet[][], boolean isFirst)
	{  
		ArrayList<Item> CandidateSet = new ArrayList<Item>();
		
		if(isFirst)
		{
			for(int i = 0; i < DataSet.length; i ++)
			{
				for(int j = 0; j < DataSet[i].length; j ++)
				{
					Item item = new Item(DataSet[i][j], 1);
					
					if(CandidateSet.contains(item))
					{
						CandidateSet.get(CandidateSet.indexOf(item)).SupportCount ++;
					}
					else
					{
						CandidateSet.add(item);
					}
				}
			}
		}
		else
		{	
			for(int i = 0; i < FrequentItemSet.size()-1; i ++)
			{
				for(int j = i+1; j < FrequentItemSet.size(); j ++)
				{
					ArrayList<String> list = Union(StrToList(FrequentItemSet.get(i).Element), 
							                       StrToList(FrequentItemSet.get(j).Element));
					
					if(list.size() == FrequentItemSet.get(0).Element.length+1)
					{	
						String [] UnionStr = list.toArray(new String[list.size()]);
						
						if(!CandidateSet.contains(new Item(UnionStr)))
						{
							int SupCnt = CalSupCnt(DataSet, UnionStr);
							
							CandidateSet.add(new Item(UnionStr, SupCnt));
						}
					}
				}
			}
		}
		
		return CandidateSet;
	}
	
	//产生频繁项集
	public ArrayList<Item> CreateFrequentItemSet(ArrayList<Item> CandidateSet, int MinSupCnt)
	{   
		for(int i = 0; i < CandidateSet.size(); i ++)
		{
			if(CandidateSet.get(i).SupportCount < MinSupCnt)
			{
				CandidateSet.remove(i--);  //注意remove后i不能自增,故i--
			}
		}

		return CandidateSet;
	}
	
	public int CalSupCnt(String[][] DataSet, String[] Str)  //计算支持度计数
	{
		int SupCnt = 0;
		int sum;
		int i, j, k;
		boolean isFound;
		
		for(i = 0; i < DataSet.length; i ++)
		{
			for(j = 0, sum = 0, isFound = true; j < Str.length && isFound; j ++)
			{
				isFound = false;
				
				for(k = 0; k < DataSet[i].length; k ++)
				{
					if(DataSet[i][k].equals(Str[j]))
					{
						sum ++;
						
                        isFound = true;  break;
					}
				}
			}
			
			if(sum == Str.length)
			{
				SupCnt ++;
			}
		}
		
		return SupCnt;
	}
	
	public ArrayList<String> StrToList(String[] DataSet) //String加入到ArrayList
	{   
		ArrayList<String> list = new ArrayList<String>();
		
		for(int i = 0; i < DataSet.length; i ++)
		{
			list.add(DataSet[i]);
		}
		
		return list;
	}
	
	//求并集
	public ArrayList<String> Union(ArrayList<String> list1, ArrayList<String> list2)  
	{   
		list1.removeAll(list2);  //ArrayList的retaiaAll方法是求交集的
		list1.addAll(list2);
		
		return list1;
	}
	
	public void PrintFrequentItemSet(ArrayList<Item> FrequentItemSet)  //输出频繁项集
	{
		System.out.println("最大频繁项集:");
		
		for(int i = 0; i < FrequentItemSet.size(); i ++)
		{
			ArrayList<String> list = StrToList(FrequentItemSet.get(i).Element);
			
			Collections.sort(list);
			
			System.out.println(list.toString() + "\tSupportCount = " + FrequentItemSet.get(i).SupportCount);
		}
	}
	
	public void PrintAssociationRules(String[] AssociationRules)  //输出关联规则
	{
		System.out.println("\n强关联规则:");
		
		for(int i = 0; i < AssociationRules.length; i ++)
		{
			System.out.println(AssociationRules[i]);
		}
	}
    
	//得到最大频繁项集
	@SuppressWarnings("unchecked")
	public ArrayList<Item> GetMaxFrequentItemSet(String[][] DataSet, int MinSupCnt)  
	{   
		boolean isFirst = true;
		ArrayList<Item> CandidateSet = new ArrayList<Item>();
		ArrayList<Item> FrequentItemSet = new ArrayList<Item>();
		ArrayList<Item> MaxFrequentItemSet = new ArrayList<Item>();
		
		do
		{
			if(!isFirst)  MaxFrequentItemSet = (ArrayList<Item>)FrequentItemSet.clone();
	
			CandidateSet = CreateCandidateSet(FrequentItemSet, DataSet, isFirst);
	
			FrequentItemSet = CreateFrequentItemSet(CandidateSet, MinSupCnt);
			
			isFirst = false;
			
		}while(!FrequentItemSet.isEmpty());
		
		return MaxFrequentItemSet;
	}
	
	//产生关联规则
	public String[] CreateAssociationRules(ArrayList<Item> MaxFrequentItemSet, String[][] DataSet, double MinConf)
	{
		ArrayList<String> list = new ArrayList<String>();
		
		for(int i = 0; i < MaxFrequentItemSet.size(); i ++)
		{
			int SupCnt = MaxFrequentItemSet.get(i).SupportCount;
			String[] Element = MaxFrequentItemSet.get(i).Element;
			
			for(int j = 1; j < (1 << Element.length)-1; j ++)  //遍历所有非空真子集
			{
				ArrayList<String> list1 = new ArrayList<String>();
				ArrayList<String> list2 = new ArrayList<String>();
				
				for(int k = 0; k < Element.length; k ++)
				{
					if((j >> k & 1) == 1)
					{
						list1.add(Element[k]);
					}
					else
					{
						list2.add(Element[k]);
					}
				}
				
				String[] Str = list1.toArray(new String[list1.size()]);
				int SubSupCnt = CalSupCnt(DataSet, Str);
				double Conf = (double)SupCnt / SubSupCnt;
				
				if(Conf >= MinConf)
				{
					DecimalFormat DecFmt = new DecimalFormat("#"); 
					
					Collections.sort(list1);
					Collections.sort(list2);
					
					String Rule = list1.toString() + " ==> " + list2.toString() 
					+ "\tConfidence = " + String.valueOf(DecFmt.format(100*Conf)) + "%";
					
					list.add(Rule);
				}
			}
		}
		
		return list.toArray(new String[list.size()]);
	}

	public static void main(String[] args) 
	{
		//数据集
		String DataSet[][] = {  
			{"I1", "I2", "I5"},
			{"I2", "I4"},
			{"I2", "I3"},
			{"I1", "I2", "I4"},
			{"I1", "I3"},
			{"I2", "I3"},
			{"I1", "I3"},
			{"I1", "I2", "I3", "I5"},
			{"I1", "I2", "I3"}};
		
		int MinSupCnt = 2;  //最小支持度计数
		double MinConf = 0.3;  //最小可信度
		
		Apriori apriori = new Apriori();
		
		ArrayList<Item> MaxFrequentItemSet = apriori.GetMaxFrequentItemSet(DataSet, MinSupCnt);
		
		apriori.PrintFrequentItemSet(MaxFrequentItemSet);
		
		String[] AssociationRules = apriori.CreateAssociationRules(MaxFrequentItemSet, DataSet, MinConf);
		
		apriori.PrintAssociationRules(AssociationRules);
	}
}


import java.util.ArrayList;
import java.util.Collections;

class Item{

	public String[] Element;
	public int SupportCount;

	public Item(String[] Element, int SupportCount)
	{
		this.Element = new String[Element.length];
		
		this.Element = Element;
		this.SupportCount = SupportCount;
	}
	
	public Item(String[] Element)
	{
		this.Element = new String[Element.length];
		
		this.Element = Element;
		this.SupportCount = 0;
	}

	public Item(String Element, int SupportCount)
	{
		this.Element = new String[1];
		
		this.Element[0] = Element;
		this.SupportCount = SupportCount;
	}
	
	public boolean equals(Object obj)  //重载equals, 用于ArrayList的contains和indexOf方法
	{
		Item item = (Item)obj;
		
		if(item.Element.length == 1)
		{
			return this.Element[0].equals(item.Element[0]);
		}
		
		Apriori apriori = new Apriori();
		
		ArrayList<String> list1 = apriori.StrToList(this.Element);
		ArrayList<String> list2 = apriori.StrToList(item.Element);
		
		Collections.sort(list1);
		Collections.sort(list2);
		
		return list1.equals(list2);
	}
}

运行结果如下图所示：