Penggunaan penggugusan subtraktif bagi menjana peraturan kabur

• Main Authors: Priyono, Agus, Ridwan, Purnomo, Alias, Ahmad Jais, Rahmat, Riza Atiq, Hassan, Azmi, Mohd Ali, Mohd Alaudin
• Format: Article
• Language: English
• Published: Universiti Kebangsaan Malaysia 2005
• Subjects: QA76 Computer software; T Technology (General)
• Online Access: http://irep.iium.edu.my/37326/1/Jurnal_Kejuruteraan_2005.pdf; http://irep.iium.edu.my/37326/; http://www.ukm.my/jkukm/v2/ojs/index.php/jkukm/article/view/127
• Institution: Universiti Islam Antarabangsa Malaysia

Sebelum ini banyak kaedah bagi menghasilkan peraturan lojik kabur telah dikaji. Lojik kabur dikenal kerana ia mampu menawarkan cara yang sederhana bagi mentakrifkan penyelesaian maklumat input yang samar-samar, kabur, bising, tidak tepat atau hilang. Model lojik kabur merupakan model berasaskan empirik, bergantung kepada pengalaman operator berbanding kefahaman mereka terhadap sistem. Disebabkan oleh operasi berasaskan peraturan, loiik kabur mempunyai kemampuan untuk menxproses sebanyak mungkin maklumat input yang munasabah serta ia mampu untuk menjana apa sahaja output yang dikehendaki. Namun penakrifon pangkalan peraturan akan cepat menjadi rumit sekiranya terlalu banyak input serta output yang dipilih. Bergantung kepada sistem, penilaian setiap kemungkinan penggabungan input mungkin tidak diperlukan sekiranya ia iarang-iarang atau sama sekali tidak boleh berlaku. Dengan menggunakan algoritma penggugusan kabur; fungsi keahlian dapat dikira berdasarkan kepada dua kemungkinan kaedah penggugusan. Pertama, gususan diunjurkan secara ortogonal ke dalam axis pemalar antiseden, dan fungsi-fungsi keahlian disesuaikan dengan unjuran tersebut. Kaedah yang kedua pula menggunakan fungsi keahlian antiseden multi dimensi, sepertimana gugusan kabur ianya diunjurkan ke dalam ruang input. Idea asas dalam penulisan kertas keria ini ialah untuk mempelajari dan menjanakan peraturan optimum yang diperlukan bagi mengawal input tanpa mengurangkan mutu kawalan. Kertas keria ini mencadangkan penggunaan pengugusanan subtraktif bagi menjana peraturan-peraturan lojii kabur pada sistem lojik kabur Takagi-Sugeno-Kang (TSK). Lojik kabur yang dicadangkan merupakan kaedah kepintaran buatan yang digunapakai pada sistem kawalan lalu lintas bandar teragih pintar Kaedah ini digabungkan dengan jaringan neural dan algoritma genetik bagi menentukan pemasaan isyarat dan ofset pada kawalan persimpangan rangkaian di Bandar Baru Bangi. Dari kajian didapati bahawa sistem mampu menghasilkan 8 pusat gugusan pada 30 (3 x 10) nilai data pada jejari gugusan 0.3 dan ia iuga menghasilkan 4 pusat gugusan pada iejari 0.5 dengan purata MSE 0.005. (Many methodologies to develop fuzzy logic rules have been previously studied. A fuzzy logic is well known because of its ability to offer a moderatemethod to translate the fuzzy, noise, inaccurate or lost input. The fuzzy logic model is based on the emphirical method depending on the operator experience comparing his understanding towards the system. According to the operation rule-based, fuzzy logic was able to process the information input immediately and also able to generate the necessary output. However, defining the rule-based quickly, becomes complex if too many input and output are chosen. Depending on the system, the assessment of each possibility input might be not necessary if this very seldom or never occur. By using the fuzzy clustering algorithm, membership function could be counted based on two possible clustering methods. First, fuzzy clustering method performed in the orthogonal axis manne4 the multivariable membership function can be projected to one-dimensional fuzzy sets. The second method is by using antecedent multi dimension membership function similar to the fuzzy cluster performed into input area. The basic idea in this paper work is how lo learn and generate the optimum rules that required controlling input without decreasing the control quality. The subtractive clustering method to generate fuzzy logic rules on Takagi-Sugeno-Kang (TSK) fuzzy system have been utilized in this study. The suggested fuzzy logic is a smart technique which is applied into urban smart-traffic. This technique combined with neural network and genetic algorithm to determine the signal timing and offset time at Bandar Baru Bangi traffic junction control system. Based on this study, it is found that the system was able to generate 8 cluster center on 30 (3 x 10) data value at 0.3 cluster radius and also able to generate 4 cluster center at 0.5 radit.rs with average MSE of 0.005).