Klasifikasi Dokumen Tumbuhan Obat Berbasis Famili Menggunakan Probalistic Neural Network
View/ Open
Date
2013Author
Muchlis, Achmad
Herdiyeni, Yeni
Zuhud, Ervizal AM
Metadata
Show full item recordAbstract
This research presents a system for the extraction of features and classification of the document medicinal plants using the chi-square method and neural network classification probalistic. Automation family identification by grouping characteristics medicinal plants contained in the document. In this research, the phase consists of collecting documents, doing conversion from hardcopy documents to softcopy into XML format, pre-process the document, featuring selection using the chi-square, document classification using PNN and evaluation using Confusion Matrix. Classification results are influenced by the number of family identifier words. Number of words identifier of a family affected by the same number of words in each document training. The more same words in each document training, the greater probability of the word being said identifier. In this research, identifier of a family does not describe the characteristics of plant taxonomy. This is due to the same number of words in each document has not practically described the characteristics of plant taxonomy. The evaluation research of classification algorithm Probalistic Neural Network (PNN) with weight value 1 and layer patterns using feature extraction using the chi-square values of alpha (α) value of 0.1 overall result of classification is 82.14%. This research represents a medical plant document system that can be used for automatic identification of families according to the taxonomy of plants.This system is useful to help users especially researchers and taxonomists in the identification document through labeling family automatically on each document, with the existing system, therefoe the users dont need to bring a book or a guide book to identify the field. Furthermore, it can overcome the limitations of people's knowledge of the diversity and the use of medicinal plants. That is because there are a lot of documents to train the same word in each document which is not characteristic of plant taxonomy.
Collections
- UT - Computer Science [2255]