ABSTRAKSI: Ancaman pornografi terhadap anak-anak dan remaja saat ini semakin mengkhawatirkan seiring derasnya perkembangan teknologi informasi. Para orangtua sebagai sosok yang paling berperan dalam proses tumbuh kembang anak mulai melakukan antisipasi guna mencegah adiksi pornografi. Namun tidak banyak orangtua mampu melakukannya karena ketersediaan waktu yang terbatas akibat kesibukan rutinitas kerja.
Tugas akhir ini bertujuan untuk membuat suatu sistem yang mampu mendeteksi suatu dokumen berbahasa Inggris terlarang yang berisi kata pornografi pada citra digital hasil scan dokumen teks. Rangkaian proses yang dilakukan oleh sistem antara lain input gambar, preprocessing, segmentasi kata, normalisasi hasil segmentasi, ekstraksi ciri, dan jaringan syaraf tiruan backpropagation.
Untuk meningkatkan performansi sistem, maka dilakukan pengujian terhadap sistem. Pengujian dilakukan dengan melakukan analisa terhadap jenis ekstraksi ciri dan beberapa parameter JST Backpropagation. Parameter yang menghasilkan akurasi maksimal yaitu ciri mean 4x4, jumlah hidden layer 1, jumlah neuron tiap layer 200, nilai learning rate 0.4, fungsi aktivasi tansig untuk hidden layer, fungsi aktivasi purelin untuk output layer, algoritma pembelajaran trainrp, dan nilai validasi 0.01. Akurasi maksimal yang diperoleh adalah 99.75% dengan waktu komputasi ± 148.72408 detik/halaman penuh.Kata Kunci : deteksi pornografi, dokumen teks, JST backpropagationABSTRACT: Pornography issues around the children and the teenagers are become worrying along with the expansion of information technology which has been growing rapidly during the day. Parents who have the duty of the children’s growth have been doing anticipations to prevent the addiction of pornography to their children. Unfortunately, not many parents can do those things according to the limited time they have.
This final project aims to build the system that can detect the forbidden English document which contains the pornography words inside from the digital image of the scanning document. The flowing processes in the system are input image, preprocessing, words segmentation, normalization, feature extraction, and the backpropagation neural network.
To improve the performance of the system, then the system is tested. The testing system is done by create an analysis of some parameters. Based on the result of performance testing system, it is known that the performance of the system reaches the highest accuracy when the feature extraction is the feature of mean from 4x4 matrix, one hidden layer, 200 neuron in every layer, the learning rate is 0.4, the activation function of hidden layer is tansig, the activation function of output layer is purelin, the learning algorithm is trainrp, and the validation value is 0.01. The accuracy that is obtained by the system is 99.75% and computation time of the system is ± 148.72408 seconds/full page.Keyword: pornography detection, text document, backpropagation neural network