项目作者: mariazm

项目描述 :
Public code of the ML course
高级语言: Python
项目地址: git://github.com/mariazm/DS_MachineLearning.git
创建时间: 2017-06-07T19:22:16Z
项目社区:https://github.com/mariazm/DS_MachineLearning

开源协议:

下载


Public code of the research paper

Law and Norms: A Machine Learning Approach to Predicting Attitudes Towards Abortion

https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2816659

Data:

Private

Abstract

Understanding the predictors of societal attitudes has been widely investigated using individuals’ responses to surveys and polls. In this paper, we use U.S. Courts of Appeals cases, sociological and attitudinal indicators, criminal statistics, and a variety of survey data to predict societal attitudes towards abortion. We create two classification models: pro-abortion attitudes for health reasons related to the mother, fetus, or rape and pro-abortion attitudes for any other personal non-health related reason. To address high dimensionality, we employ factor analysis to group indicators. Logistic regression and random forests performed best among three types of classifiers evaluated on AUC and accuracy.

For pro abortion attitudes related to health related abortions, the most important factor contains sociological and attitudinal indicators about the frequency of contact with family and friends. The most important U.S. Courts of Appeals indicators include the religion, political affiliation, and ABA ratings of the judges in the Circuit pool. The most important crime indicators are the rates of violent and property crimes.

For pro abortion attitudes towards non-health related abortions, the most important factor contains sociological and attitudinal indicators about job satisfaction, religious preference, religion raised in, and beliefs about the Bible. The most important U.S. Courts of Appeals indicators include the religion and political affiliation of the judges in the Circuit panel. These legal indicators are grouped in the same factor as the most important crime indicators, which are rates of crimes against society, crimes against property, and violent crimes.

Authors:

  • Maria Leonor Zamora Maass

NYU Courant Institute, Center for Data Science

  • Luisa Eugenia Quispe Ortiz

NYU Courant Institute, Center for Data Science

  • Kristen Kwan

NYU Courant Institute, Center for Data Science

  • Advisor: Daniel L. Chen

University of Toulouse 1 - Toulouse School of Economics Institute for Advanced Studies/Harvard Law School LWP; Harvard Law School

July 31, 2016