TMU_Security_Incidents

Overview

By collecting security incident data reported by TMU's Community Safety and Security Site, this project aims to:

• Forecast monthly incidents using SARIMAX for the upcoming year.
• Dashboard all reported incident data into an interactive view.
• Provide insights which can help TMU Security plan and allocate rescources to enhance safety.
• Serve as the basis for future machine learning and NLP projects.

Packages Used

• Pandas, Selenium, BeautifulSoup4, Seaborn, statsmodels, Streamlit, Vega-Altair

Results

Streamlit Dashboard

https://tmu-security-incidents-dashboard.streamlit.app/

Preview

Monthly Forecast

Process Walkthrough

Data Cleaning & Exploration

A subset of 32 reports are only titled 'Toronto Police Service News Release':

• Incident type had to be extracted from the text found in incident_details by keyword/phrase mapping.

Location Inconsistencies:

• Ordering:

  • Yonge Street and Gould Street vs Gould Street and Yonge Street.

• Renaming:

  • TMU Image Centre vs Ryerson Image Centre.

• Level of Precision:

  • Tim Hortons at the Yonge- Dundas Square vs Yonge- Dundas Square.

• Locations were standardized by the following:

  • Streets of intersections ordered alphabetically.
  • Ryerson replaced with TMU.
  • Precision reduced to general building or area.
  • Spelling errors corrected.

Visuals

Time Series Analysis

Decomposition

Monthly Incidents + 12 Month Forecast

Work in Progress - NLP + K-means clustering

• Endcoding entities in incident_details

  Example

  'tmu community members': 'TMU_MEMBER',
  'tmu community member': 'TMU_MEMBER',
  'ryerson community members': 'TMU_MEMBER',
  'ryerson community member': 'TMU_MEMBER',
  'community members': 'TMU_MEMBER',
  'community member': 'TMU_MEMBER',
  
  'tmu community guests': 'TMU_GUEST',
  'tmu community guest': 'TMU_GUEST',
  'ryerson community guests': 'TMU_GUEST',
  'ryerson community guest': 'TMU_GUEST',
  'community guests': 'TMU_GUEST',
  'community guest': 'TMU_GUEST',
  
  'tmu security': 'TMU_SECURITY',
  'ryerson security': 'TMU_SECURITY',
  'security': 'TMU_SECURITY',
  
  'members of the public': 'MEMBER_OF_THE_PUBLIC',
  'member of the public': 'MEMBER_OF_THE_PUBLIC',
  'members of public': 'MEMBER_OF_THE_PUBLIC',
  'member of public': 'MEMBER_OF_THE_PUBLIC',
  
  'individual' : 'INDIVIDUAL',
  'individuals' : 'INDIVIDUAL',
  
  'toronto police services': 'TORONTO_POLICE_SERVICES',
  'toronto police service': 'TORONTO_POLICE_SERVICES',
  'toronto police': 'TORONTO_POLICE_SERVICES',
  ' tps ': ' TORONTO_POLICE_SERVICES ',
  'police officers': 'TORONTO_POLICE_SERVICES',
  'police officer': 'TORONTO_POLICE_SERVICES',
  'police': 'TORONTO_POLICE_SERVICES',
  'officers': 'TORONTO_POLICE_SERVICES',
  'officer': 'TORONTO_POLICE_SERVICES',
  
  'toronto paramedic services': 'TORONTO_PARAMEDIC_SERVICES',
  'toronto paramedic service': 'TORONTO_PARAMEDIC_SERVICES',
  'paramedics': 'TORONTO_PARAMEDIC_SERVICES',
  'paramedic': 'TORONTO_PARAMEDIC_SERVICES',
  

• Extract features from sentences based on Part-of-speech tagging in relation to entity

• TF-IDF

• K-means clustering