Group 4 - Lee Durrell, Brian Martin, Richard Dieckhaus, Scott Leeman


 

Introduction

Our Approach

Our Model

Conclusion

Sources

Introduction

 

Our goal is to try and predict the number and locations of murders in St. Louis City for 2013.

To this end we went through several methods and created models to create a forecast.

 

Detective Picture

Our Approach

 

•Methods Attempted

–Looked at Regression

Small sample size, overwhelming number of factors with insufficient data

 

–Sought Correlations

 

Some of the factors are included below:

Biological / Psychological

Age

Gender

Mental Illness

Personality Disorders

Ecological

Population Size

Neighborhood Conditions

Weather

Socioeconomic

High Poverty

Education Level

Occupation Level

Cultural and Societal

Gang Violence

Drug/ Alcohol Usage

Gun Ownership

Home Vacancies

 

–Looked at Linear Trend Lines

 

 

Based on previous studies, and our early attempts to produce a good model for prediction, we chose to use a simple linear model and apply exponential smoothing to forecast.

Due to the number of correlates, it is too difficult to come with a model that shows a meaningful connection and provides an accurate method of forecasting.  We strictly followed the murder data to develop our predictions.

Return to Top

Graph Image

Our Model

 

Monthly Murder Data

We used 134 months of murder data in our data set.  This includes data from 2002 to 2012, and the first two recorded months of 2013.

 

 

  January February March April May June July August September October November December Total
2002 13 11 5 10 8 4 12 9 14 15 3 9 113
2003 8 3 7 3 8 6 11 8 3 1 2 14 74
2004 4 3 12 5 8 17 2 21 4 13 16 9 114
2005 8 8 7 13 12 12 14 8 15 11 8 15 131
2006 13 3 8 12 12 8 15 12 12 6 17 11 129
2007 9 5 10 9 13 14 10 12 10 11 20 15 138
2008 9 14 7 16 17 23 11 18 21 12 9 10 167
2009 5 9 10 8 16 12 10 15 13 22 15 8 143
2010 13 8 7 14 10 10 9 7 13 13 26 14 144
2011 2 12 11 10 16 18 10 9 8 7 7 3 113
2012 11 6 6 10 12 13 12 11 8 9 6 9 113
TOTAL 95 82 90 110 132 137 116 130 121 120 129 117  
2013 15 5                      

 

 

 

Seasonal ScatterPlot of Monthly Murder Data

 

In order to perform analysis on the table, we used a pivot table in excel to create seasonal indices.

 

Seasonal Indices Pivot table

 

We then applied those seasonal indices to our data, and used exponential smoothing to create forecasts.

 

Date Murders Deseas Forecast
Jan-02 13 15.73 #N/A
Feb-02 11 15.42 15.73
Mar-02 5 6.38 15.48
Apr-02 10 10.45 8.20
May-02 8 6.96 10.00
Jun-02 4 3.36 7.57
Jul-02 12 11.89 4.20
Aug-02 9 7.96 10.35
Sep-02 14 13.30 8.43
Oct-02 15 14.36 12.32
Nov-02 3 2.67 13.96
Dec-02 9 8.84 4.93
Jan-03 8 9.68 8.06
Feb-03 3 4.20 9.35
Mar-03 7 8.94 5.23
Apr-03 3 3.13 8.20
May-03 8 6.96 4.15
Jun-03 6 5.03 6.40
Jul-03 11 10.90 5.31
Aug-03 8 7.07 9.78
Sep-03 3 2.85 7.61
Oct-03 1 0.96 3.80
Nov-03 2 1.78 1.53
Dec-03 14 13.75 1.73
Jan-04 4 4.84 11.35
Feb-04 3 4.20 6.14
Mar-04 12 15.32 4.59
Apr-04 5 5.22 13.18
May-04 8 6.96 6.81
Jun-04 17 14.26 6.93
Jul-04 2 1.98 12.79
Aug-04 21 18.56 4.14
Sep-04 4 3.80 15.68
Oct-04 13 12.45 6.18
Nov-04 16 14.25 11.19
Dec-04 9 8.84 13.64
Jan-05 8 9.68 9.80
Feb-05 8 11.21 9.70
Mar-05 7 8.94 10.91
Apr-05 13 13.58 9.33
May-05 12 10.45 12.73
Jun-05 12 10.07 10.90
Jul-05 14 13.87 10.23
Aug-05 8 7.07 13.14
Sep-05 15 14.25 8.29
Oct-05 11 10.53 13.05
Nov-05 8 7.13 11.04
Dec-05 15 14.73 7.91
Jan-06 13 15.73 13.37
Feb-06 3 4.20 15.25
Mar-06 8 10.21 6.41
Apr-06 12 12.54 9.45
May-06 12 10.45 11.92
Jun-06 8 6.71 10.74
Jul-06 15 14.86 7.52
Aug-06 12 10.61 13.39
Sep-06 12 11.40 11.16
Oct-06 6 5.75 11.35
Nov-06 17 15.14 6.87
Dec-06 11 10.80 13.49
Jan-07 9 10.89 11.34
Feb-07 5 7.01 10.98
Mar-07 10 12.77 7.80
Apr-07 9 9.40 11.78
May-07 13 11.32 9.88
Jun-07 14 11.74 11.03
Jul-07 10 9.91 11.60
Aug-07 12 10.61 10.25
Sep-07 10 9.50 10.54
Oct-07 11 10.53 9.70
Nov-07 20 17.82 10.37
Dec-07 15 14.73 16.33
Jan-08 9 10.89 15.05
Feb-08 14 19.62 11.72
Mar-08 7 8.94 18.04
Apr-08 16 16.72 10.76
May-08 17 14.80 15.52
Jun-08 23 19.29 14.94
Jul-08 11 10.90 18.42
Aug-08 18 15.91 12.40
Sep-08 21 19.94 15.21
Oct-08 12 11.49 19.00
Nov-08 9 8.02 12.99
Dec-08 10 9.82 9.01
Jan-09 5 6.05 9.66
Feb-09 9 12.61 6.77
Mar-09 10 12.77 11.44
Apr-09 8 8.36 12.50
May-09 16 13.93 9.19
Jun-09 12 10.07 12.98
Jul-09 10 9.91 10.65
Aug-09 15 13.26 10.06
Sep-09 13 12.35 12.62
Oct-09 22 21.07 12.40
Nov-09 15 13.36 19.33
Dec-09 8 7.86 14.56
Jan-10 13 15.73 9.20
Feb-10 8 11.21 14.42
Mar-10 7 8.94 11.85
Apr-10 14 14.63 9.52
May-10 10 8.71 13.60
Jun-10 10 8.39 9.69
Jul-10 9 8.92 8.65
Aug-10 7 6.19 8.86
Sep-10 13 12.35 6.72
Oct-10 13 12.45 11.22
Nov-10 26 23.16 12.20
Dec-10 14 13.75 20.97
Jan-11 2 2.42 15.19
Feb-11 12 16.82 4.97
Mar-11 11 14.05 14.45
Apr-11 10 10.45 14.13
May-11 16 13.93 11.18
Jun-11 18 15.10 13.38
Jul-11 10 9.91 14.75
Aug-11 9 7.96 10.88
Sep-11 8 7.60 8.54
Oct-11 7 6.70 7.79
Nov-11 7 6.24 6.92
Dec-11 3 2.95 6.37
Jan-12 11 13.31 3.63 2012 Actual 
Feb-12 6 8.41 11.37 113
Mar-12 6 7.66 9.00
Apr-12 10 10.45 7.93 2012 Forecasted
May-12 12 10.45 9.94 107.39
Jun-12 13 10.90 10.35
Jul-12 12 11.89 10.79
Aug-12 11 9.72 11.67
Sep-12 8 7.60 10.11
Oct-12 9 8.62 8.10
Nov-12 6 5.34 8.52
Dec-12 9 8.84 5.98
Jan-13 15 18.14 8.27
Feb-13 5 7.01 16.17
Mar-13 8.84 11.29 8.84
Apr-13 10.80 11.28 10.80
May-13 11.18 9.74 11.18
Jun-13 10.03 8.41 10.03
Jul-13 8.73 8.65 8.73
Aug-13 8.67 7.66 8.67
Sep-13 7.86 7.47 7.86
Oct-13 7.55 7.23 7.55
Nov-13 7.29 6.49 7.29 2013 Forecasted
Dec-13 6.65 6.54 6.65 112.04

Our final forecast for total murders in St. Louis City in 2013 is 112.

 

We also must try to break down murder by location. We chose to do this by predicting for each district in St. Louis City.

District Map

District Map

 

We did this prediciton based on percentage of annual murders per district over the last 11 years.

Predictions of Murders by District

Destrict predictions

 

 

Return to Top

Conclusions

Conclusion

 

After looking at several methods, due to the number of correlates, data availability, and sample size we decided to go with a simple model that tracks trends in murder based solely on the murder data itself.

To predict location of murders we simply applied the historical percentage of murders by district to each district and our totalpredicted number of murders for the year.

 

 

Prediction Quote

 

Return to Top

Documentation

Sources

https://www.ncjrs.gov/pdffiles1/nij/grants/211973.pdf

http://www.kmov.com/news/editors-pick/St-Louis-police--195189401.html

http://forprin.dev.zoe.co.nz/files/pdf/Gorr_Olligschalger_and_Thompson,_Short-term.pdf

http://research.stlouisfed.org/fred2/series/MOSSURN

http://cad.sagepub.com/content/49/3/339.full.pdf+html

http://www.drtomoconnor.com/1010/1010lect01a.htm

http://www.euronews.com/2013/02/25/forecasting-crime/

http://bjs.ojp.usdoj.gov/content/pub/pdf/ics.pdf


http://www.nytimes.com/2009/06/19/nyregion/19murder.html?pagewanted=all&_r=0 


http://ajp.psychiatryonline.org/article.aspx?articleID=172630

 

 

 

 

Return to Top