Data Management project @ ISB - Crime data of Chicago city

CRIME DATA OF

CHICAGO CITY

METHODOLOGY USED

• Most of the variable were categorical in nature which are to aggregated to perform the clustering analysis .

• The data is then uploaded into the database and aggregated by DISTRICT , WARD and COMMUNITY AREA

to find the count of crimes by categories, following which the hierarchal clustering is performed

• Tools used :

• TERADATA

• TABLEAU

• SPOTFIRE

CLUSTERING THE CRIME TYPES

CLUSTER-1

SUM_LIQUOR_LAW_VIOLATION

SUM_INTIMIDATION

SUM_PUBLIC_PEACE_VIOLION

CLUSTER-4

SUM_NON_CRIMINAL_SUB_SPECI

CLUSTER-2

SUM_ARSON

SUM_KIDNAPPING

SUM_OTHER_OFFENSE

SUM_CRIMINAL_DAMAGE

SUM_BURGLARY

SUM_SEX_OFFENSE

SUM_OFFENSE_INVOL_CHLDRN

SUM_MOTOR_VEHICLE_THEFT

SUM_OBSCENITY

CLUSTER-3

SUM_NARCOTICS

SUM_PROSTITUTION

SUM_ASSAULT

SUM_WITH_PUBLIC_OFFICER

SUM_ROBBERY

SUM_GAMBLING

SUM_CRIM_SEXUAL_ASSAULT

SUM_HOMICIDE

SUM_BATTERY

SUM_INTER_WITH_PUB_OFFICER

SUM_WEAPONS_VIOLATION

CLUSTER-5

SUM_CRIMINAL_TRESPASS

SUM_STALKING

SUM_OTHER_NARCOTIC_VIOLATION

SUM_PUBLIC_INDECENCY

SUM_DECEP_PRACT

SUM_THEFT

SUM_NON_CRIMINAL

INSIGHT: WHERE MALE TO FEMALE RATIO IS LESS, THERE ARE MORE CRIMES COMMITTED.

MAP SHOWING MALE TO FEMALE RATIOBUBBLE SIZE SHOWS SUM OF TOTAL NUMBER OF CRIMES COMMITTED

MAP SHOWING DISTRIBUTION OF THE HOUSE HOLD INCOME

INSIGHT: WHERE HOUSE HOLD INCOME IS LOW, CRIMES ARE MORE

MAP SHOWING DISTRIBUTION OF THE PER CAPITA INCOME

INSIGHT: WHERE PER CAPITA IS MORE, MORE THEFTS WERE COMMITTED

MAP SHOWING MALE TO FEMALE RATIOBUBBLE SIZE SHOWS SUM OF HOMICIDES WITH DISTRICT NUMBERS

INSIGHT: WHERE MALE TO FEMALE RATIO IS LESS, HOMICIDES ARE MORE

MAP SHOWING MALE TO FEMALE RATIOBUBBLE SIZE SHOWS SUM OF CRIME - PROSTITUTION.

INSIGHT: WHERE MALE TO FEMALE RATIO IS LESS, PROSTITUTION IS MORE

HEAT MAP – DISTRICT

INSIGHT: ON CLUSTERING THE COUNT OF CRIMES, WE OBSERVE THAT DISTRICTS ARE GETTING CLUSTERED INTO 3 CATEGORIES –HIGH, INTERMEDIATE AND LOW. AREA

INSIGHT: WE SEE THAT FEW DISTRICTS IN THE TOP TWO CLUSTERS HAVE INVERSE PATTERN ON FEW VARIABLES.

PARALLEL COORDINATE – DISTRICT

THE SAME PATTERN CAN BE OBSERVED FOR WARD AND COMMUNITY AREA

FUTURE SCOPE OF STUDY

• Application of Dummy Variables can be explored

• Association rules among crime types can be applied

• Location type based clustering can be performed

• Network analysis – To identify (closeness )distance between two crimes

APPENDICES

HEAT MAP – COMMUNITY AREA

HEAT MAP – WARD

PARALLEL COORDINATE – COMMUNITY AREA

PARALLEL COORDINATE– WARD