ASCUE 51ST Annual Conference June 2018 Predicting Effective Internships: Developing a Text Analysis Based Algorithm Dr. Brian Hoyt (hoyt@ohio.edu) Ohio University
Best Internship Profile (BIP) Predicts skill development as dependent variable Predicts Return on Internship Investment (ROII) as dependent variable
Predicting BIP Applications Partner focus/impact – selection of projects; recruitment and selection of interns; allocation of resources for internship/program Student focus/impact – skill development focus for selection; project experience focus for selection; pre- internship preparation Campus focus/impact – all majors (and undecided) skill development emphasis (critical thinking/problem solving, data analysis)
Modeling Approach Determine core “footprint” factors as independent variables Patterns (i.e. plan structure and goal setting; partner involvement in planning; types of partners/industry sectors) Clusters (i.e. project types; prerequisites completed; ROII focus) Determine weighting Statistical significance Inferential statistics – mean differences, factorial design, correlation relationships
Modeling Approach Internship data sources Intern characteristics Partner characteristics Project Description Learning Plan WIP postings Project deliverables Evaluations (mid, self, skills) Activity Reports Recorded post internship simulated interviews Final report
Best Internship Profile Index Predictor Intern skill profile index Internship project profile index
Software Utilization SPSS – footprint and weighting Text analysis - construct development and impact measures vagueness conscientiousness and extroversion influencers emotional intelligence / Jones intensity model Provalis Research’s WordStat7 and QDA miner.
Big Data and New Data Streams New constructs New characteristics (i.e. personality inventories and assessments) Available data streams
Abstract: Effective internships should be measured based on the skill development demonstrated by student interns and the return on internship investment (ROII) for intern partners. An ongoing challenge is controlling factors associated with the planning and implementation of internships that are more likely to result in an effective internship. This presentation will review a data analytics approach to developing a predictive model to be used identify the most important characteristics of an effective internship. The algorithm uses text analysis methodologies and software to construct a “Best Internship Profile”. The presentation reviews how the model strengthens internship planning, implementation, and intern reporting processes and tools required to maximize intern skill development and ROII for partners. The “Best Internship Profile” model used data collected from over 500 internships. Data streams used in the text analysis approach were collected from internship planning documents, learning plans, weekly activity reports, in process work outputs, partner evaluations, and final report documents. Data analysis software used is IBM SPSS Statistics 23 and text analysis software included Provalis Research’s WordStat7 and QDA miner.