Please use this identifier to cite or link to this item:
https://ptsldigital.ukm.my/jspui/handle/123456789/513511Full metadata record
| DC Field | Value | Language |
|---|---|---|
| dc.contributor.advisor | Azuraliza Abu Bakar, Professor Dr. | |
| dc.contributor.author | Anis Suhailis binti Abdul Kadir (P53569) | |
| dc.date.accessioned | 2023-10-16T04:37:28Z | - |
| dc.date.available | 2023-10-16T04:37:28Z | - |
| dc.date.issued | 2014-07-11 | |
| dc.identifier.other | ukmvital:79976 | |
| dc.identifier.uri | https://ptsldigital.ukm.my/jspui/handle/123456789/513511 | - |
| dc.description | Perlombongan petua sekutuan terdiri daripada petua sekutuan positif dan petua sekutuan negatif. Kajian lepas menunjukkan perlombongan petua sekutuan memberi tumpuan kepada petua sekutuan positif yang dihasilkan daripada set item kerap dengan hubungan positif. Pengetahuan daripada set item positif kerap adalah lazim dan boleh jangka. Bagaimana pun, beberapa kajian telah membuktikan bahawa petua sekutuan negatif berpotensi diteroka dan menyumbang kepada pengetahuan yang menarik kerana pengetahuan yang dihasilkan sukar untuk dijangka dan unik. Sehubungan itu, objektif kajian ini adalah untuk meneroka pendekatan berbeza daripada perlombongan petua sekutuan negatif sedia ada dengan memperkenalkan set item negatif kerap bersama dengan set item positif kerap. Set item negatif kerap berlaku apabila item atau set item tidak hadir dalam transaksi melepasi nilai ambang yang ditetapkan. Kajian ini mencadangkan pendekatan set item Positif dan Negatif Kerap (PNK) untuk algoritma model peramalan. Kajian ini mempunyai dua fasa. Fasa 1, pentakrifan pendekatan PNK dan algoritma PNK dicadangkan untuk membina dua model peramalan iaitu model pengesanan sisihan dan model pengelasan sekutuan. Fasa 2, algoritma PNK yang dicadang di Fasa 1, digunakan keatas satu kajian kes iaitu pengesanan pencerobohan rangkaian. Dalam Fasa 1, satu model untuk pengesanan sisihan berasaskan algoritma PNK dipanggil PNK-Pengesanan Sisihan (PNK-PS) dicadangkan untuk menjana set item positif dan negatif kerap. Satu pengukuran darjah sisihan untuk PNK diformulasi, dikenali sebagai Darjah Sisihan PNK. Algoritma apriori diubahsuai untuk mengira set item negatif kerap. Ia termasuk kekuatan kolektif sebagai satu pengukuran pemangkasan untuk mengira kekuatan set item. Kemudian, satu lagi model berasaskan PNK dicadangkan untuk pengelasan sekutuan yang dipanggil PNK-Pengelasan Sekutuan (PNK-KS). PNK-KS menjana petua sekutuan positif dan negatif untuk pengelasan. Sepuluh set data domain umum dipilih untuk ujikaji kedua-dua model peramalan. Pengesanan sisihan diukur menggunakan dua pengukuran tanda aras iaitu nisbah tertinggi dan nisbah terlitup. Hasil ujikaji menunjukkan yang PNK-PS berprestasi lebih baik secara signifikan dari dua kaedah pengesanan sisihan yang lain. Sementara itu PNK-KS berprestasi lebih baik berbanding empat kaedah yang lain. Dalam Fasa 2, model PNK-PS dan model PNK-KS dikaji lebih lanjut keatas data trafik rangkaian untuk mengesan pencerobohan rangkaian. Hasil ujikaji menunjukkan PNK boleh digunakan dalam domain ini iaitu model PNK-PS berupaya mengesan pencerobohan rangkaian dengan kadar pengesanan yang tinggi. Manakala model PNK-KS berjaya mengelaskan data rangkaian dengan tepat berbanding kaedah lain. Algoritma berasaskan PNK menunjukkan yang pengetahuan yang penting boleh diekstrak daripada petua negatif dan positif, dan bermakna dalam masalah tertentu.,Association rules mining consists of positive and negative association rules. Literature shows that association rule mining research focuses on the positive association rules, which are generated from frequent itemsets with positive relationships. The knowledge from frequent positive itemsets is common and predictable. However, several literatures exhibit that negative association rule is potentially to be explored and contribute to interesting knowledge, which is unpredictable and unique. Therefore, the objective of this study is to explore the different approaches of negative association rule mining by introducing frequent negative itemsets together with frequent positive itemsets. The frequent negative itemsets occurs when an item or itemset is not present in the transaction and exceeded the specified threshold value. This study proposes the Frequent Positive and Negative itemsets (FPN) approach for predictive model algorithms. The study has two phases. Phase 1, the definition of FPN approach together with the based algorithms is proposed to build two predictive models namely outlier detection model and associative classification model. Phase 2, the FPN algorithms proposed in Phase 1 are employed in real world case study, network intrusion detection. In Phase 1, an algorithm for outlier detection based on FPN called FPN-Outlier Detection (FPN-OD) is proposed to generate frequent positive and negative itemsets. A new outlier degree measure for FPN is formulated, known as FPN Outlier Degree. The Apriori algorithm is modified to compute the frequent negative itemsets. It includes the collective strength as a pruning measures in order to calculate the interestingness of itemsets. Then, another FPN based algorithm is proposed for the associative classification which is called the FPN-Associative Classification (FPN-AC). FPN-AC generates the positive and negative associative rules for classification. Ten sets of public domain data are chosen for the experiment for both predictive models. The detection of outliers is measured using two benchmark measurements which are top ratio and coverage ratio. The experimental results show that the FPN-OD performed significantly better than the other two outlier detection methods. Meanwhile, FPN-AC is measured in term of accuracy of classification. The result indicates that FPN-AC performs better than the other four associative classifiers. In Phase 2, the proposed FPN based algorithms are employed in real world data, network traffic data for intrusion detection. The FPN algorithms are further explored for building the predictive models on network traffic data for intrusion detection. The experimental results showed that the FPN can be used in this domains. FPN-OD is capable to detect network intrusion with higher detection rate compared to other methods. The FPN based algorithms showed that important knowledge can be extracted from the positive and negative rules, and meaningful in certain problems.,PhD | |
| dc.language.iso | eng | |
| dc.publisher | UKM, Bangi | |
| dc.relation | Faculty of Information Science and Technology / Fakulti Teknologi dan Sains Maklumat | |
| dc.rights | UKM | |
| dc.subject | Model pengelasan sekutuan | |
| dc.subject | Data mining | |
| dc.title | Pendekatan set item positif dan negatif kerap untuk model pengelasan sekutuan dan pengesanan sisihan | |
| dc.type | Theses | |
| dc.format.pages | 170 | |
| dc.identifier.callno | QA76.9.D343A556 2014 3 tesis | |
| dc.identifier.barcode | 001074 | |
| Appears in Collections: | Faculty of Information Science and Technology / Fakulti Teknologi dan Sains Maklumat | |
Files in This Item:
| File | Description | Size | Format | |
|---|---|---|---|---|
| ukmvital_79976+SOURCE1+SOURCE1.0.PDF Restricted Access | 2.46 MB | Adobe PDF |  View/Open |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.