Saturday, April 26, 2014

Crime Prediction Through Twitter

Twitter universe apparently storing a myriad of information that proved useful for uncovering crime , even before the crime was committed . Researchers at the University of Virginia showed that of a tweet we can predict crimes that may occur when applying the correct analysis .

A research paper published in the scientific journal Decision Support Systems last month said that the analysis of tweets by putting location information can be useful in predicting 19 to 25 types of crimes , especially for offenses such as stalking , theft , and several other types of attacks . " The results were quite surprising . Though people rarely tweet about crime directly , "said lead researcher Matthew Gerber of Predictive Technology Lab at the University of Viginia . Gerber said the tweet even not directly related to the crime may contain information about related activities .
( Read: cara membuat blog sendiri )

" People often tweet about their daily activities . Routines that brought them into an environment where crime might occur , " said Gerber . " So if I'm going to go get drunk tweet tonight , and many people also planned so, then we know there a number of crimes related to the things that generate crime . Indirectly . " For this study , Gerber and his colleagues analyzed the tweets of the city of Chicago for a particular environment ( square kilometers ) and city crime data base . Then they were able to predict where certain crimes are likely to occur in the area , something that could assist in the deployment of police resources .

" This approach allows analysts to quickly visualize and identify areas with a high concentration of historical crime . Crime will come usually occurs in the vicinity of the crime in the past . Mapping hot -spot into a valuable predictor of crime . " In recent years , the idea of ​​" predictive policing " has become a discourse , police departments rely on the analysis of data from companies such as IBM . It is one of the few other studies that show how tweets can be analyzed to predict elections , epidemics , and other important events .

Gerber said the data is relatively easy to use Twitter because tweets publicly available, and many that include location information . In addition to the researchers themselves , do not need to go directly to the troubled area just learn the information.

The study was funded by the U.S. Army , which used the same technique to determine the threat in places like Iraq and Afghanistan . Gerber noted there are limitations . Adequate historical data are absolutely necessary , and some types of crimes such as kidnapping and arson , may not fit into the same pattern of predictability . The reason , the researchers themselves could not explain .

But the New York police department have contacted Gerber and he has begun reviewing the data from the city to determine whether the results can be replicated from Chicago . He hopes to be able to utilize information from other social media to see if they could help improve prediction . Its main purpose is to determine whether this technique could be used in a very practical way . " We are not saying that this will reduce the level of crime , it is the next step , " said Gerber . ( Nis )


No comments:

Post a Comment