Academic analytics : comparative study of decision tree in predicting success in the licensure examination of graduates / by Jeffrey Arsenal Clarin.
By: Clarin, Jeffrey Arsenal [author.]
Language: English Description: xi, 60 leaves : color illustrations ; 28 cmContent type: text Media type: unmediated Carrier type: volumeSubject(s): Teaching -- Examinations, questions, etc | Teachers -- Examinations, questions, etcGenre/Form: Academic theses.DDC classification: 371.26 Dissertation note: Thesis (Doctor in Information Technology) -- Cebu Institute of Technology - University, Cebu City, 2020. Abstract: The researcher aimed to compare the similarities and dissimilarities between the researcher-developed academic analytics utilizing the CART and C4.5 algorithms when utilized to predict success in the licensure examination of graduates. This study adapted the knowledge discovery in databases (KDD) systems development method to identify the valid and valuable patterns in the gathered data as well as the comparative research methodology to identify similarities and differences between entities. The researcher utilized 348 instances coming from the 2012 to 2017 (5years) teacher education graduates in Kalibo, Aklan with the following attributes: Entrance Exam Result, Specialization Subjects; Professional Education Subjects; General Education Subjects, LET Review, and LET Exam Result. Among the similarities found were: both CART and C4.5 generated a decision tree utilizing the 4 steps advocated by Lewis (2000), both CART & C4.5 utilized the same parameters established for the training data sets, and, both CART & C4.5 had a strong agreement or are both reliable, as established by Kappa statistic. Among the dissimilarities found in this study were: a) the overall predictive capability (reliability) of the classification tree model of CART was established to be 94.2029%; whereas the summary of the evaluation of the training set for C4.5 was found to be 91.6667%; b) the confusion matrix for CART showed an overall accuracy of 90.80%; for C4.5, 91.67%; and, c) the Kappa value for CART was established to be 0.7965; for C4.5, 0.8195. Statistical analysis established a t-value of 2.03 and p-value of 0.180. In the event that the p-value was more than the 0.05 Alpha confidence level set for this study, the hypothesis that there is no significant difference between the constructed dissimilarities observed in CART and C4.5 decision tree algorithms when utilized in predicting the graduates’ success in the licensure examination was accepted.| Item type | Current location | Home library | Call number | Status | Date due | Barcode | Item holds |
|---|---|---|---|---|---|---|---|
THESIS / DISSERTATION
|
COLLEGE LIBRARY | GRADUATE LIBRARY Filipiniana | 371.26 C544 2020 (Browse shelf) | Not for loan |
Thesis (Doctor in Information Technology) -- Cebu Institute of Technology - University, Cebu City, 2020.
Includes bibliographical references.
The researcher aimed to compare the similarities and dissimilarities between the researcher-developed academic analytics utilizing the CART and C4.5 algorithms when utilized to predict success in the licensure examination of graduates.
This study adapted the knowledge discovery in databases (KDD) systems development method to identify the valid and valuable patterns in the gathered data as well as the comparative research methodology to identify similarities and differences between entities.
The researcher utilized 348 instances coming from the 2012 to 2017 (5years) teacher education graduates in Kalibo, Aklan with the following attributes: Entrance Exam Result, Specialization Subjects; Professional Education Subjects; General Education Subjects, LET Review, and LET Exam Result.
Among the similarities found were: both CART and C4.5 generated a decision tree utilizing the 4 steps advocated by Lewis (2000), both CART & C4.5 utilized the same parameters established for the training data sets, and, both CART & C4.5 had a strong agreement or are both reliable, as established by Kappa statistic.
Among the dissimilarities found in this study were: a) the overall predictive capability (reliability) of the classification tree model of CART was established to be 94.2029%; whereas the summary of the evaluation of the training set for C4.5 was found to be 91.6667%; b) the confusion matrix for CART showed an overall accuracy of 90.80%; for C4.5, 91.67%; and, c) the Kappa value for CART was established to be 0.7965; for C4.5, 0.8195.
Statistical analysis established a t-value of 2.03 and p-value of 0.180. In the event that the p-value was more than the 0.05 Alpha confidence level set for this study, the hypothesis that there is no significant difference between the constructed dissimilarities observed in CART and C4.5 decision tree algorithms when utilized in predicting the graduates’ success in the licensure examination was accepted.
There are no comments for this item.