带频繁区域的空间并置模式挖掘方法

打开文本图片集

关键词：空间并置模式挖掘；频繁区域；候选区域；拓展区域；区域粗参与度中图分类号：TP391 文献标志码：A 文章编号：1001-3695（2025）07-022-2086-10doi：10.19734/j. issn.1001-3695. 2024.10.0456

Abstract：Thisstudy focusedonspatialco-location patern mining，aiming toexplore theco-locationrelationships between spatialfeatures.Whiletraditionalmethodscanidentifyfrequentlyco-locationpatterns，theycannotdeterminethespecificspatialregions wherethese pattersoccur.Toaddress thisissue，thisstudyproposedanovel spatialco-locationpattrmining algorithmwithfrequentregions.Thealgorithmwasdividedintotwostages：thefirststageusedanagglomerativehierarchical clustering method topartition thespacebasedonthedatacharacteristics，andthenconfirmedthe proximityrelationships between instances within each cluster.Thesecond stage introduced theconcepts ofco-location patern presence regionsand regionalparticipationdegree，ndbasedonthese，itincrementallidentifiedthefrequentregionsofco-locationpaterns.To acceleratetheidentificationoffrequent regionsandthepattmmining process，thealgorithmquicklyconstructedcandidate regionsforhigher-orderpatternsbyexpanding theregionsofsub-paternsandusedrough participation degres tofilterout candidateregionsthatwereunlikelytobefrequent inadvance.Finally，extensive experiments onrealandsyntheticdatasets havedemonstratedthepeformanceoftheproposedalgorithmintermsof thenumberof spatialco-locationpaternswith frequentregions generated，theaccuracyoffrequentregions，andtheprecisionoffrequentregions.Onreal datasets，theaccuracyof thealgorithmrangesbetweenO.83andO.95.Furthermore，inexperiments evaluating thescalabilityofthealgorithm， whenthenumberoffeaturesinthedataset ismoderate，theperformanceof thePROC-Colalgorithmisapproximatelytwiceas fast as the current state-of-the-art multi-level algorithm.

Key words：spatialco-location patern mining；frequentregions；candidateregions；expandedregions；rough regional participation index

0 引言

空间并置模式挖掘1自2001年提出以来，取得了显著进展。（剩余25237字）