2015 Homicide Prediction For Saint Louis City Shashin

  • Slides: 19
Download presentation
2015 Homicide Prediction For Saint Louis City Shashin Amatya Yi Gao Lauren Reuther INFSYS-6833

2015 Homicide Prediction For Saint Louis City Shashin Amatya Yi Gao Lauren Reuther INFSYS-6833 Group B Homicide

Agenda • • Our Objectives Problems Background Our Approach Analysis Prediction Improving Prediction Recommendation

Agenda • • Our Objectives Problems Background Our Approach Analysis Prediction Improving Prediction Recommendation Homicide

Objectives • Accurately predict homicide in St. Louis City • Help effectively allocate required

Objectives • Accurately predict homicide in St. Louis City • Help effectively allocate required resources in the area Homicide

Problems Limited datasets Homicide numbers don’t always match Hard to obtain variable data for

Problems Limited datasets Homicide numbers don’t always match Hard to obtain variable data for period of time Gang Activity Drug Related Violence Hard to obtain Historical Data Yearly Population Density Yearly Income by Neighborhood Yearly Unemployment Rate by Neighborhood Yearly Percentage of People with Kids Yearly Percentage of Races by Neighborhood Homicide

Brief Background Total Population: 318, 563 2013 Total Homicide: 120 2014 Total Homicide: 157

Brief Background Total Population: 318, 563 2013 Total Homicide: 120 2014 Total Homicide: 157 2014 - 5 th Most dangerous city Violent crime per 100, 000 inhabitants: 1, 594 Poverty Rate: 26. 6% Homicide

Our Approach/Methods We looked at following methods Linear Regression Logarithmic Regression Polynomial Regression Three

Our Approach/Methods We looked at following methods Linear Regression Logarithmic Regression Polynomial Regression Three Month Weighted Moving Average Best Approach Three Month Weighted Moving Average Homicide

Potential Contributors to Homicide Following variables might influence homicide rate Population Density Median Household

Potential Contributors to Homicide Following variables might influence homicide rate Population Density Median Household Income Median Home Value Unemployment Rate Race/Ethnicity Education Level Median Age Marital Status Male/Female Ratio Median Home Value Homicide

Why Best Approach Used only one substantive variable and time index Forecasts solely based

Why Best Approach Used only one substantive variable and time index Forecasts solely based on historical patterns of data Works better with limited datasets Other variables can be ignored Time trend, seasonal adjustment, cyclical component, and irregular component are already included Homicide

How This Method Is Used By Distributing Weight 50% to the month of 2014

How This Method Is Used By Distributing Weight 50% to the month of 2014 33% to the month of 2013 17% to the month of 2012 Formula Used following formula to calculate the value Weighted Average = ((3*2014 month) + (2*2013 month) + (1*2012 month)/6 Homicide

Our Prediction - 2015 140 120 120 104 100 80 80 60 60 95

Our Prediction - 2015 140 120 120 104 100 80 80 60 60 95 44 35 40 20 0 18 21 20 Central North 2014 South 2015 Round Up Total Predicted Homicide 35 40 0 18 35 12 Central North 2014 South 2015 Round Down Total Predicted Homicide

Our Prediction - 2015 Year Month 2015 Jan 2015 Feb 2015 Mar 2015 Apr

Our Prediction - 2015 Year Month 2015 Jan 2015 Feb 2015 Mar 2015 Apr 2015 May 2015 Jun 2015 Jul 2015 Aug 2015 Sep 2015 Oct 2015 Nov 2015 Dec Totals Corridor North South Central 8 6 0 3 3 3 5 5 2 9 1 1 6 3 2 5 5 1 9 5 2 11 3 2 10 3 1 13 3 3 10 3 2 15 4 2 104 44 21 Total Predicted Homicides in 2015 169 Homicide

Projected Homicide (All Corridors) Homicide

Projected Homicide (All Corridors) Homicide

Central Corridor Homicide – (2012 -2015) Homicide

Central Corridor Homicide – (2012 -2015) Homicide

Location of Homicides Occurring Year 2015 2015 2015 Corridor Month North South Central Jan

Location of Homicides Occurring Year 2015 2015 2015 Corridor Month North South Central Jan 8 6 0 Feb 3 3 3 Mar 5 5 2 Apr 9 1 1 May 6 3 2 Jun 5 5 1 Jul 9 5 2 Aug 11 3 2 Sep 10 3 1 Oct 13 3 3 Nov 10 3 2 Dec 15 4 2 Totals 104 44 21 Homicides are more likely to occur in the North Corridor. Homicide

Weather vs Number of Homicides Central - 2012 North Central - 2013 Central -

Weather vs Number of Homicides Central - 2012 North Central - 2013 Central - 2014 North - 2012 North - 2014 r be m ce De ve m be r r be to No Oc em be r st Se pt Au gu Ju ly e ay South - 2012 Ju n M M r be m De ce be r r be ve m No r Oc be to st em Se pt Au gu Ju ly e Ju n ril ay M ar M Ap ch b Fe Ja n 0 ril 1 Ap 2 ar 3 b 4 Fe 5 ch 18 16 14 12 10 8 6 4 2 0 Ja n 6 North - 2013 Weather South - 2013 South - 2014 Weather - 2013 Weather - 2014 r be m ce De ve m be r r be to No r Oc be em st Se pt gu Ju ly e Ju n ril ay M Au pt Se Ap em be r Oc to be r No ve m be r De ce m be r st gu Au Ju ly e Ju n ay M ril Ap ch ar M b Fe Ja n 0 ch 5 ar 10 M 15 Fe 20 b 100 90 80 70 60 50 40 30 20 10 0 Ja n 25 Weather - 2012 Homicide

How to Improve Prediction Providing adequate historical data Properly defining gang and drug related

How to Improve Prediction Providing adequate historical data Properly defining gang and drug related activities Providing accurate data for variables Exploring other modeling methods Properly testing model Readjusting model if needed

Recommendation Properly reallocating active duty personnel Neighborhood Crime Watch Job creation Deploy programs for

Recommendation Properly reallocating active duty personnel Neighborhood Crime Watch Job creation Deploy programs for citizen awareness Deploy programs to uplift neighborhood situation Homicide

Questions? Homicide

Questions? Homicide

Citations • http: //www. slmpd. org/Crimereports. shtml • http: //www. areavibes. com/st. +louis-mo/neighborhoods/academy/areavibe/ •

Citations • http: //www. slmpd. org/Crimereports. shtml • http: //www. areavibes. com/st. +louis-mo/neighborhoods/academy/areavibe/ • https: //www. stlouis-mo. gov/neighborhoods/ • http: //www. point 2 homes. com/US/Neighborhood/MO/St-Louis/Central. Corridor-Demographics. html#Population • http: //www. weather. gov/climate/local_data. php? wfo=lsx