Title

Using Decision Trees to Analyze Online Learning Data

Location

Kansas State University-Manhattan Campus

Session Type

Presentation

Streaming Media

Session Abstract

In machine learning, decision trees enable researchers to identify possible indicators (variables) that are important in predicting classifications, and these offer a sequence of nuanced groupings. For example, are there “tells” which would suggest that a particular student will achieve a particular grade in a course? Are there indicators that would identify learners who would select a particular field of study vs. another?

This session will introduce how decision trees are used to model data based on supervised machine learning (with labeled training set data) and how such models may be evaluated for accuracy with test data, with the open-source tool, RapidMiner. Several related analytical data visualizations will be shared: 2D spatial maps, decision trees, and others. Attendees will also experience how 2x2 contingency tables work with Type 1 and Type 2 errors (and how the accuracy of the machine learning model may be assessed) to represent model accuracy, and the strengths and weaknesses of decision trees applied to some use cases from higher education. In this session, various examples of possible outcomes will be discussed and related pre-modeling theorizing (vs. post-hoc) about what may be seen in terms of particular variables. The basic data structure for running the decision tree algorithm will be described. If time allows, relevant parameters for a decision tree model will be discussed: criterion (gain_ratio, information_gain, gini_index, and accuracy), minimal size for split, minimal leaf size, minimal gain, maximal depth (based on the need for human readability of decision trees), confidence, and pre-pruning (and the desired level).

Creative Commons License

Creative Commons Attribution-Noncommercial 4.0 License
This work is licensed under a Creative Commons Attribution-Noncommercial-Share Alike 4.0 License.

Share

COinS
 

Using Decision Trees to Analyze Online Learning Data

Kansas State University-Manhattan Campus

In machine learning, decision trees enable researchers to identify possible indicators (variables) that are important in predicting classifications, and these offer a sequence of nuanced groupings. For example, are there “tells” which would suggest that a particular student will achieve a particular grade in a course? Are there indicators that would identify learners who would select a particular field of study vs. another?

This session will introduce how decision trees are used to model data based on supervised machine learning (with labeled training set data) and how such models may be evaluated for accuracy with test data, with the open-source tool, RapidMiner. Several related analytical data visualizations will be shared: 2D spatial maps, decision trees, and others. Attendees will also experience how 2x2 contingency tables work with Type 1 and Type 2 errors (and how the accuracy of the machine learning model may be assessed) to represent model accuracy, and the strengths and weaknesses of decision trees applied to some use cases from higher education. In this session, various examples of possible outcomes will be discussed and related pre-modeling theorizing (vs. post-hoc) about what may be seen in terms of particular variables. The basic data structure for running the decision tree algorithm will be described. If time allows, relevant parameters for a decision tree model will be discussed: criterion (gain_ratio, information_gain, gini_index, and accuracy), minimal size for split, minimal leaf size, minimal gain, maximal depth (based on the need for human readability of decision trees), confidence, and pre-pruning (and the desired level).

http://newprairiepress.org/isitl/2017/Presentations/7