Developing Machine Learning Classifiers on Uncovering Patterns of Intimate Partner Violence Risk

A 2019-2020 Seed Grant Award
Principal Investigator: Ambika Krishnakumar
Co-Investigator: Rachel Razza

With Bei Yu (iSchool) co-I, and Ying Zhang (HDFS Doctoral Student).
CUSE Grant – Innovative & Interdisciplinary Research, $20,000.

We seek support from the CUSE’s Innovative and Interdisciplinary Research Grant committee to conduct a two-year study that will utilize text-mining algorithms to uncover patterns or typologies of intimate partner violence (IPV) from women’s narratives. For the purpose of this grant, IPV is conceptualized as the motivational need of aggressors to exert coercive control in close relationships and aggressors need to use physical, sexual, and psychological violence to reinforce coercive control in relationships.

Moving away from employing closed-ended questionnaires to study IPV, we aim to:

  1. collect a bank of IPV narratives from women’s online discussion forums which will be coded and classified based on information about coercive control and violent behaviors,
  2. enter the predefined text data into the computer so as to allow the machine to learn from the text and generate classification algorithms about IPV typologies based on comparative unique-word frequencies and word combinations, and
  3. develop an interactive website based on information gained from machine learning algorithms and statistical models.

This interdisciplinary study will incorporate information from family science, information science, and computational linguistic science. We anticipate that the interactive website could be utilized by women experiencing IPV to comprehend the nature of their violent relationships, and by first responders and others (e.g., counseling centers) to make informed decisions when dealing with violent situations. We will use data from this project to seek additional funding for the development of a “Mobile App” that could be used by institutions to better serve IPV victims.