POLICE TRY TO PREDICT CITY CRIME HOT SPOTS.

Byline: Ryan Oliver Staff Writer

The Los Angeles Police Department has allocated $80,000 to develop a computer program that officials hope will use crime statistics and socioeconomic data to forecast crime before it occurs.

Commanders say they could use the forecast to deploy resources and take other precautions before a crime wave actually hits.

``There is nothing really serious like this that has been done before with any consistency, and certainly nothing on the scale as large at this department,'' said LAPD Assistant Chief George Gascon, who is spearheading the project. ``To that extent, we're excited to be on the frontier of a new way of police work.''

Gascon said such a system may sound ambitious but the project is realistic and could be online within a few years.

The LAPD has already adopted the Compstat program that Chief William Bratton developed when he headed the New York Police Department. Gascon said Compstat could be used as the basis for the new program.

``There is no reason computer forecasting can't work,'' he said. ``It's really nothing more than an analysis of probabilities.''

But one of the nation's leading experts in the emerging field of geographic crime forecasting said that, despite several attempts, no one has been able to develop a program with the accuracy that the LAPD is hoping for.

``It's going to be a really blunt tool,'' said Wilpen Gorr, a management systems researcher at Carnegie Mellon University in Pittsburgh. ``I don't think it'll ever get really sharp.''

Gorr recently led two studies funded by $400,000 in federal grants to develop crime-forecasting models for 10-square-block areas of Pittsburgh and Rochester, N.Y. Researchers collected criminal reports, history, trends and 911 calls and studied seasonal variations to form the basis of their data pool.

But when Gorr compared the predictions with real crime numbers, the forecasts were often too far off to be useful.

``I had some expectations you could do these 10-block areas,'' he said, ``I thought they were ideal. It was really disappointing when we ran our complete data and it didn't hold up.''

The forecasts were more accurate if plotted on a 20-by-20 block area. But such an area would likely be too large for law enforcement to crack down effectively on any potential hot spots.

Gorr said it's more difficult to forecast for smaller areas because they have fewer criminal incidents, so the computer has fewer data points to analyze.

The LAPD's Gascon remains optimistic about crime forecasting, however, saying the department will devise its own methods that could prove more accurate than Carnegie Mellon's. But even if it doesn't, it's worth the risk, he said.

``We recognize this is experimental work for us, and when you're trying new things there's the possibility it won't work as anticipated,'' he said. ``I think we have to try it for the Los Angeles experience.''

Gascon also wants the computer program to be able to analyze the backgrounds of suspects to see if their backgrounds and criminal records are similar to those who have gone on to commit violent crimes. The computer would flag that individual and Gascon would like social workers to try to intervene.

Ryan Oliver, [phone]

ryan.oliver(at)dailynews.com