REU Site: Data Science of Risk and Human Activity

印第安纳大学、普渡大学、印第安纳波利斯大学(IUPUI)的本科生研究体验(REU)项目将为来自美国各地的八名本科生提供开展人类活动数据科学研究的机会。学生们将在暑期花十周时间与IUPUI的教职员工合作，研究与犯罪、冲突、政治不稳定和日常活动相关的预测建模和风险评估项目。这些学生还将参加数据科学训练营，作为该项目的一部分，该项目将提供数据科学基础(统计学、机器学习和软件开发)方面的培训。数据科学是一个快速发展的领域，因为产生的数据量和种类都在增加。无论是在工业界还是学术界，拥有数学、统计、计算、数据分析和数据建模等方面技能的毕业生都很受欢迎。REU网站将提高本科生在STEM领域攻读研究生学位的意识、准备和兴趣，在STEM领域，数据科学正成为一个更大的重点。选择在毕业后从事工业职业的学生也将为应对越来越多的公司的数据挑战做好更好的准备，在这些公司中，数据科学是高度优先的。REU项目解决了风险和人类活动数据科学方面的新挑战。在问题领域1工作的学生将专注于学习如何对时空犯罪预测的问题进行排序。将开发根据风险对犯罪热点进行排名的机器学习模型，该模型将针对犯罪学中出现的基于排名的损失函数而量身定做。第二个项目将侧重于从移动传感器时间序列中对活动类型进行分类。学生们将把现有的技术与深度学习算法进行比较，深度学习算法使用的是专门为加速度计和陀螺仪的3D数据开发的体系结构。第三个研究领域将侧重于不同申诉数据的点过程模型。特别是，学生将开发撒哈拉以南非洲冲突数据的时空模型，并结合同一时间段的地理位置相关推文。

This Research Experiences for Undergraduates (REU) program at Indiana University Purdue University Indianapolis (IUPUI) will provide eight undergraduate students from across the United States with the opportunity to conduct research on the data science of human activity. The students will spend ten weeks during the summer working with IUPUI faculty on projects related to predictive modeling and estimation of risk with applications to crime, conflict, political instability and daily routine activity. The students will also attend a data science bootcamp as part of the program that will provide training in the foundations of data science (statistics, machine learning, and software development). Data science is a rapidly growing field due to increases in the volume and variety of data being generated. Graduates with skill sets at the intersection of mathematics, statistics, computing, data analysis, and data modeling are in high demand, both in industry and academia. The REU site will increase undergraduate student awareness, preparation, and interest in pursuing graduate degrees in STEM fields where data science is becoming a larger focus. Students choosing to pursue a career in industry upon graduation will also be better prepared to meet the data challenges of the increasing number of companies where data science is a high priority. The REU projects address new challenges in the data science of risk and human activity. Students working in problem area 1 will focus on learning to rank problems for space-time crime prediction. Machine learning models for ranking crime hotspots according to risk will be developed that will be tailored to ranking based loss functions that arise in criminology. The second project will focus on classifying types of activity from mobile sensor time series. The students will compare existing techniques to deep learning algorithms employing architectures developed specifically for 3d data from the accelerometer and gyroscope. The third area of research will focus on point process models of heterogeneous grievance data. In particular, students will develop spatio-temporal models for conflict data in Sub-Saharan Africa coupled with geolocated, relevant Tweets from the same time period.