A Survey of Data Mining Techniques on MedicalData for Finding Temporally Frequent Diseases

Data mining techniques have potential to discover hiddenrelationships in the data of medical databases. This will helpin understanding the prevailing situations in healthcaredomain with respect to patients, their medical conditions andtreatments. Medical databases are very bulky that needcomputerized programs to find latent trends that will help inmedical diagnosis and treatment. In the wake of data miningtechniques, especially medical data mining techniques, thehealth care domain has made significant progress in usingthe technologies in prevention and diagnosis of disease.With respect to data mining techniques, the traditionalfrequent pattern discovering techniques [1], [2], [3], [4], [5],[6], [7], [8] are not sufficient to know the temporal nature ofdiseases. These techniques do not consider the elapsed timebetween two events and thus cannot produce valuableinsights into temporally frequent diseases as they do not takethe time dimension as variable in their framework.Spenceley and Warren [9] explored temporal data miningwith respect to taking intelligent inputs to an online medicalapplication.Catley, Stratti, and McGregor [10] emergingtechniques related to temporal data mining on medical timeseries data sets.Catley et al. [11] have extended their worklater with respect to multi-dimensional medicaldata.Meamarzadeh, Khayyambashi and Saraee [12] appliedtemporal data mining techniques that helped in discoveringhidden relationships in medical data sets.Shuxia and Zheng[13] proposed fuzziness approach for mining interminacytemporal data.Tsumoto, Hirano and Iwata [14] appliedtemporal data mining for characterization of medicalpractice.Adaptive fuzzy cognitive maps are used by Froelichand Wakulicz-Deja [15] for mining temporal data in medicaldata sets.Berlingerio, Bonchi, Giannotti and Turini [16]believed that clinical databases contain temporal data thatcan be exploited to discover intelligence that supports inmaking decisions pertaining to patients’ health anddiagnosis.Abe, Yokoi, Ohsaki and Yamaguchi [17] proposedan integrated environment for medical data mining.Tsumotoand Hirano [18] explored the mining possible trajectoriesfrom medical data sets. More details about all theseresearches can be found in section II.Our contributions in this paper include the analysis of stateof-the-art of the existing data mining techniques that areused for temporal data mining on medical data besidessummarizing and providing future directions of the research.This remainder of this paper is structured as follows. SectionII reviews related literature. Section III summarizes thefindings pertaining to extracting temporally frequentdiseases while section IV concludes the paper.