{
    "article_id": 549,
    "public_id": "f2a2633c-d224-4e07-bf8a-831514b3c8f1",
    "internal_identifier": null,
    "source_identifier": "5:196",
    "status": "published",
    "type": "research_article",
    "title": "Comparison of Text Classification Performances in Balanced and Imbalanced Datasets",
    "title_secondary": null,
    "abstract": "Text mining, which aims to derive unknown and useful information from available textual data, is one of the subfield\r\nof data mining. Text mining transforms unstructured textual data into structured form by utilizing various methods. Data mining\r\ntechniques can also be applied to textual data as a result of transformation to structured form. Text classification which is one of\r\nthe widely favoured subject of text mining is the process of assigning texts into predefined classes. As a result of this,\r\nclassification is realized more rapidly and consistently.\r\n\r\nIn this study, text classification performances in balanced and imbalanced datasets were compared. Corpora which consist of\r\nTurkish and English texts were utilized. The features of 4 datasets including 5 classes in each were as follows: Corpus 1 and\r\nCorpus 3 include Turkish contents and Corpus 2 and Corpus 4 include English contents. 3375 training and 1125 test documents\r\nwere included in the Corpus 1 and Corpus 2 which are balanced datasets, whereas 1825 training and 825 test documents were\r\nincluded in the Corpus 3 and Corpus 4 which are imbalanced datasets.\r\n\r\nDocuments included in datasets were pre-processed, weighted as binary and tf-idf and document vectors were obtained. kNN\r\nwas preferred for text classification and Manhattan Distance, Harmonic Mean, Inner Product, Squared Chord and Dice’s\r\nCoefficient were selected for the measurement of similarities of document vectors.\r\n\r\nResults were taken into consideration in terms of classification success as well as process time. Since the number of the documents within the classes are equal, it is seen that more successful classification was obtained in terms of average values in both Turkish and English content balanced datasets but the process time was longer.  ",
    "language": "en",
    "english_title": null,
    "english_abstract": null,
    "doi": null,
    "doi_url": null,
    "resource_url": "https://www.set-science.com/a/f2a2633c-d224-4e07-bf8a-831514b3c8f1",
    "pdf_url": "https://www.set-science.com/a/f2a2633c-d224-4e07-bf8a-831514b3c8f1/pdf",
    "og_image": "https://www.set-science.com/assets/img/og-card-v2.png",
    "authors": [
        {
            "public_id": "08ce8960-f0e7-4aaa-ad6f-becdec80d948",
            "given": "Mehmet Fatih",
            "family": "Karaca",
            "full": "Mehmet Fatih Karaca",
            "literal": null,
            "orcid": null,
            "orcid_url": null,
            "email": "mfkaraca@gmail.com",
            "corresponding": true,
            "sequence": "first",
            "affiliations": [
                {
                    "name": "Department of Computer Technologies, Gaziosmanpaşa University",
                    "department": null,
                    "country": null,
                    "country_code": null,
                    "ror": null,
                    "ror_url": null
                }
            ]
        },
        {
            "public_id": "17fb48f2-0552-4c87-b100-953c21bdcc83",
            "given": "Şafak",
            "family": "Bayır",
            "full": "Şafak Bayır",
            "literal": null,
            "orcid": null,
            "orcid_url": null,
            "email": "safakbayir@karabuk.edu.tr",
            "corresponding": false,
            "sequence": "additional",
            "affiliations": [
                {
                    "name": "Department of Computer Engineering, Karabük University",
                    "department": null,
                    "country": null,
                    "country_code": null,
                    "ror": null,
                    "ror_url": null
                }
            ]
        }
    ],
    "keywords": [
        "text mining, text classification, kNN, balanced dataset, imbalanced dataset"
    ],
    "translations": [
        {
            "language": "en",
            "title": "Comparison of Text Classification Performances in Balanced and Imbalanced Datasets",
            "abstract": "Text mining, which aims to derive unknown and useful information from available textual data, is one of the subfield\r\nof data mining. Text mining transforms unstructured textual data into structured form by utilizing various methods. Data mining\r\ntechniques can also be applied to textual data as a result of transformation to structured form. Text classification which is one of\r\nthe widely favoured subject of text mining is the process of assigning texts into predefined classes. As a result of this,\r\nclassification is realized more rapidly and consistently.\r\n\r\nIn this study, text classification performances in balanced and imbalanced datasets were compared. Corpora which consist of\r\nTurkish and English texts were utilized. The features of 4 datasets including 5 classes in each were as follows: Corpus 1 and\r\nCorpus 3 include Turkish contents and Corpus 2 and Corpus 4 include English contents. 3375 training and 1125 test documents\r\nwere included in the Corpus 1 and Corpus 2 which are balanced datasets, whereas 1825 training and 825 test documents were\r\nincluded in the Corpus 3 and Corpus 4 which are imbalanced datasets.\r\n\r\nDocuments included in datasets were pre-processed, weighted as binary and tf-idf and document vectors were obtained. kNN\r\nwas preferred for text classification and Manhattan Distance, Harmonic Mean, Inner Product, Squared Chord and Dice’s\r\nCoefficient were selected for the measurement of similarities of document vectors.\r\n\r\nResults were taken into consideration in terms of classification success as well as process time. Since the number of the documents within the classes are equal, it is seen that more successful classification was obtained in terms of average values in both Turkish and English content balanced datasets but the process time was longer.  \r\n",
            "keywords": [
                "text mining, text classification, kNN, balanced dataset, imbalanced dataset"
            ]
        }
    ],
    "subjects": [],
    "subject_names": [],
    "published_date": "2018-06-23",
    "published_year": "2018",
    "received_date": null,
    "accepted_date": null,
    "volume": "2",
    "issue": null,
    "page_first": "223",
    "page_last": "223",
    "article_number": null,
    "conference": {
        "short_name": "ISAS 2018",
        "full_title": "Ist International Symposium on Innovative Approaches in Scientific Studies",
        "name": "Ist International Symposium on Innovative Approaches in Scientific Studies",
        "acronym": "ISAS 2018",
        "location": "Kemer-Antalya, Türkiye",
        "start_date": "2018-04-11",
        "end_date": null,
        "public_id": "1d94fec9-c4fe-4034-9157-0b205eef4e56"
    },
    "proceedings": {
        "title": "Ist International Symposium on Innovative Approaches in Scientific Studies",
        "series_title": "Setsci conference indexing system",
        "publisher": "SET Technology",
        "isbn": null,
        "eisbn": null,
        "issn": "2618-6446",
        "eissn": null,
        "doi": null,
        "doi_prefix": null,
        "slug": "setsci-volume-2-2018-legacy-5",
        "public_id": "03da7b1c-cb6e-4f3b-aee4-ac41ee60fd75",
        "url": "https://www.set-science.com/p/setsci-volume-2-2018-legacy-5"
    },
    "license": {
        "name": "CC BY 4.0",
        "url": "https://creativecommons.org/licenses/by/4.0/",
        "code": "CC-BY-4.0",
        "open_access": true
    },
    "copyright": {
        "holder": "SET Technology",
        "year": "2018"
    },
    "funding_statement": null,
    "funders": [],
    "references": [],
    "statements": {
        "acknowledgements": null,
        "conflict_of_interest": null,
        "data_availability": null,
        "ai_disclosure": null
    },
    "credit": [],
    "history": {
        "published": "2018-06-23"
    }
}