Hi,
I am looking to do a forecasting across space & time in predixion.
Is it possible to do the same when we have to predict occurrences across both time and location ?

I think it is possible using R but using some different packages. But if we get a R solution , can we put it back to Predixion though it is not a R ARIMA or R Auto ARIMA .

Please help

Vidya

Hi Vidya!

Thank you very much for posting your question to the Forum!

I will address the second question first. With the integration between R and Predixion Insight 3.2 we now support a wider range of Modeling Algorithms. As an advanced user you can modify the R templates to customize these algorithms. Here are some additional details for this support – http://predixionsoftware.com/Insigh…ration.pdf

Can you tell me more about the Space-Time problem you are trying to solve?

Warm regards,
Jeff Willis

Thanks Jeff and sorry for delay in reply.
I think this documentation provided will help me in understanding the intergation surely.

On the space-time problem, I was thinking of predicting the crime using the existing and history data. This requires predicting both the place and the time.

Thanks in anticipation
Vidya

Hi Vidya,

I am a data scientist at Predixion. I would like to join this conversation and see if I can help.

To better understand the problem, I would like to know more of your data, like what is your target. You mentioned ARIMA in previous posts, does that mean you are predicting the counts of certain crimes in certain location? Do you have any other variables that contributes to the crimes, like population in that area, education level etc?If you could provide some sample data, I can help on how to formulate the problem and how to handle it in Predixion.

For your spatial mining problem, here are some references .
http://www.spatial.cs.umn.edu/sdm.html
http://www.innovativegis.com/basis/…opic28.htm

Thank you,
Jerry

Hi Jerry,
I am looking more from understanding and analyzing if we can use predixion for a real big problem like predicting crime.
As I have data of Chicago which has time and space (location) , I was interested how to handle this in Predixion , now that R is integrated.

https://data.cityofchicago.org/Public-Safety/Crimes-One-year-prior-to-present/x2n5-8w5q

Thanks in advance.

Vidya

Hi Vidya,

The data looks very nice. I will work on it and see what interesting thing I can find there. Then I will schedule a call with you to go through the results.

Best,
Jerry

Thanks Jerry,
Will look forward for our call

Vidya

Dear Vidya!

Thank you for your time today to have a meeting with Jerry and myself. I just wanted to update this forum post with details we discussed. Jerry explained that you can only predict “what type of crime” will happen (from city of Chicago crime dataset) however you cannot predict “when crime will happen” because there is no information about non-crime situations. This crime dataset includes “different type of crimes” thus is more meaningful for classification purposes and Predixion Insight Classify method is better to go with.

Warm Regards,

Dhaval Maru
Predixion Software