Application of Multi Agent System for Information Extraction Needs Blood on Twitter using Naive Bayes Classifier

Abstract

Increased demand sometimes not accompanied by the increasing number of donors, the UTD PMI shortage of blood supply, thus requiring the donor's blood is searching needs. Along the development era of social media used Twitter to share information needs blood. The sentence contained in the tweet is not a sentence is structured, so it must use a computational approach to word processing on a tweet. Text processing is used to change the unstructured text into structured text, which includes stages case folding, filtering, tokenizing, hose replacement and stop word removal. Nae Bayes classifier method is applied using an agent technology to perform tweet data classification that relevant to the information of blood needs, so that the information about blood needs from tweet can spread in real time and also be able to work autonomous. We use Prometheus methodology to designing an agent. Agents that built consisting of a collection agent, preprocessing agent, classification agents and matching agent. From the test results to classify the agent system shows that the value of the smallest accuracy of 89.1% is generated in the test with the first sample, and for the highest accuracy value of 96.1% generated in the third test using 920 training data.