MATLAB数据分析与挖掘实战案例 (13).zip

function queue=evaluateAttr(queue,data,attrIndex,targetValue,minSupportCount,minImprovement,labelStateIndex) %% 评价第i列属性的某个值是否有潜力加入队列 % 判断是否有足够的数据 numInstances = size(data,1); if numInstances<=minSupportCount return; end % 根据第i列进行数据排序 [~,b_index] = sort(data(:,attrIndex)); data = data(b_index,:); % 排序后的数据 targetLeft =0; targetRight = sum(data(:,end)==labelStateIndex); % 初始化参数 bestMerit = 0.0; bestSplit = 0.0; bestSupport = 0.0; bestSubsetSize = 0; % lessThan = true; lessThan =1; % denominators leftCount = 0; rightCount = numInstances; %% for 遍历 for i=1:numInstances inst = data(i,:); if inst(1,end) == labelStateIndex targetLeft=targetLeft+1; targetRight=targetRight-1; end leftCount=leftCount+1; rightCount=rightCount-1; % move to the end of any ties if i < numInstances data_ii = data(i+1,:); if data_ii(1,attrIndex)==inst(1,attrIndex) continue; end end % evaluate split if targetLeft >= minSupportCount delta = (targetLeft / leftCount) - bestMerit; if delta > 0 bestMerit = targetLeft / leftCount; % bestSplit = inst[attrIndex]; bestSplit = inst(1,attrIndex); bestSupport = targetLeft; bestSubsetSize = leftCount; % lessThan = true; lessThan =1; else if delta == 0 % break ties in favour of higher support if targetLeft > bestSupport bestMerit = targetLeft / leftCount; % bestSplit = inst[attrIndex]; bestSplit = inst(1,attrIndex); bestSupport = targetLeft; bestSubsetSize = leftCount; % lessThan = true; lessThan =1; end end end end if targetRight >= minSupportCount delta =(targetRight / rightCount) - bestMerit; if delta > 0 bestMerit = targetRight / rightCount; % bestSplit = inst[attrIndex]; bestSplit =inst(1,attrIndex); bestSupport = targetRight; bestSubsetSize = rightCount; % lessThan = false; lessThan =0; else if delta == 0 if targetRight > bestSupport bestMerit = targetRight / rightCount; % bestSplit = inst[attrIndex]; bestSplit = inst(1,attrIndex); bestSupport = targetRight; bestSubsetSize = rightCount; % lessThan = false; lessThan =0; end end end end end % for delta = bestMerit- targetValue; % Have we found a candidate split? if bestSupport > 0 && delta / targetValue >= minImprovement % AttrStateSup ass = new AttrStateSup(attrIndex,(float)bestSplit, % (int)bestSubsetSize,(int)bestSupport,lessThan); % 构造结构体，插入queue queue_node =struct('attrIndex',attrIndex,'stateIndex',bestSplit,... 'stateCount', bestSupport,'allCount',bestSubsetSize,'support',... bestSupport/bestSubsetSize,'lessThan',lessThan); % pq.add(ass); queue= queue_push(queue,queue_node); end end