Dataset Information

Enhancing data pipelines for forecasting student performance: integrating feature selection with cross-validation.

ABSTRACT: Educators seek to harness knowledge from educational corpora to improve student performance outcomes. Although prior studies have compared the efficacy of data mining methods (DMMs) in pipelines for forecasting student success, less work has focused on identifying a set of relevant features prior to model development and quantifying the stability of feature selection techniques. Pinpointing a subset of pertinent features can (1) reduce the number of variables that need to be managed by stakeholders, (2) make "black-box" algorithms more interpretable, and (3) provide greater guidance for faculty to implement targeted interventions. To that end, we introduce a methodology integrating feature selection with cross-validation and rank each feature on subsets of the training corpus. This modified pipeline was applied to forecast the performance of 3225 students in a baccalaureate science course using a set of 57 features, four DMMs, and four filter feature selection techniques. Correlation Attribute Evaluation (CAE) and Fisher's Scoring Algorithm (FSA) achieved significantly higher Area Under the Curve (AUC) values for logistic regression (LR) and elastic net regression (GLMNET), compared to when this pipeline step was omitted. Relief Attribute Evaluation (RAE) was highly unstable and produced models with the poorest prediction performance. Borda's method identified grade point average, number of credits taken, and performance on concept inventory assessments as the primary factors impacting predictions of student performance. We discuss the benefits of this approach when developing data pipelines for predictive modeling in undergraduate settings that are more interpretable and actionable for faculty and stakeholders.

Supplementary information

The online version contains supplementary material available at 10.1186/s41239-021-00279-6.

SUBMITTER: Bertolini R

PROVIDER: S-EPMC8591701 | biostudies-literature |

REPOSITORIES: biostudies-literature

ACCESS DATA

Similar Datasets

Project description:BACKGROUND:Self organizing maps (SOM) enable the straightforward portraying of high-dimensional data of large sample collections in terms of sample-specific images. The analysis of their texture provides so-called spot-clusters of co-expressed genes which require subsequent significance filtering and functional interpretation. We address feature selection in terms of the gene ranking problem and the interpretation of the obtained spot-related lists using concepts of molecular function. RESULTS:Different expression scores based either on simple fold change-measures or on regularized Student's t-statistics are applied to spot-related gene lists and compared with special emphasis on the error characteristics of microarray expression data. The spot-clusters are analyzed using different methods of gene set enrichment analysis with the focus on overexpression and/or overrepresentation of predefined sets of genes. Metagene-related overrepresentation of selected gene sets was mapped into the SOM images to assign gene function to different regions. Alternatively we estimated set-related overexpression profiles over all samples studied using a gene set enrichment score. It was also applied to the spot-clusters to generate lists of enriched gene sets. We used the tissue body index data set, a collection of expression data of human tissues as an illustrative example. We found that tissue related spots typically contain enriched populations of gene sets well corresponding to molecular processes in the respective tissues. In addition, we display special sets of housekeeping and of consistently weak and high expressed genes using SOM data filtering. CONCLUSIONS:The presented methods allow the comprehensive downstream analysis of SOM-transformed expression data in terms of cluster-related gene lists and enriched gene sets for functional interpretation. SOM clustering implies the ability to define either new gene sets using selected SOM spots or to verify and/or to amend existing ones.

Project description:Exploring the molecular mechanisms of breast cancer is essential for the early prediction, diagnosis, and treatment of cancer patients. The large scale of data obtained from the high-throughput sequencing technology makes it difficult to identify the driver mutations and a minimal optimal set of genes that are critical to the classification of cancer. In this study, we propose a novel method without any prior information to identify mutated genes associated with breast cancer. For the somatic mutation data, it is processed to a mutated matrix, from which the mutation frequency of each gene can be obtained. By setting a reasonable threshold for the mutation frequency, a mutated gene set is filtered from the mutated matrix. For the gene expression data, it is used to generate the gene expression matrix, while the mutated gene set is mapped onto the matrix to construct a co-expression profile. In the stage of feature selection, we propose a staged feature selection algorithm, using fold change, false discovery rate to select differentially expressed genes, mutual information to remove the irrelevant and redundant features, and the embedded method based on gradient boosting decision tree with Bayesian optimization to obtain an optimal model. In the stage of evaluation, we propose a weighted metric to modify the traditional accuracy to solve the sample imbalance problem. We apply the proposed method to The Cancer Genome Atlas breast cancer data and identify a mutated gene set, among which the implicated genes are oncogenes or tumor suppressors previously reported to be associated with carcinogenesis. As a comparison with the integrative network, we also perform the optimal model on the individual gene expression and the gold standard PMA50. The results show that the integrative network outperforms the gene expression and PMA50 in the average of most metrics, which indicate the effectiveness of our proposed method by integrating multiple data sources, and can discover the associated mutated genes in breast cancer.

Project description:BackgroundLong noncoding RNAs (lncRNAs) have dense linkages with a plethora of important cellular activities. lncRNAs exert functions by linking with corresponding RNA-binding proteins. Since experimental techniques to detect lncRNA-protein interactions (LPIs) are laborious and time-consuming, a few computational methods have been reported for LPI prediction. However, computation-based LPI identification methods have the following limitations: (1) Most methods were evaluated on a single dataset, and researchers may thus fail to measure their generalization ability. (2) The majority of methods were validated under cross validation on lncRNA-protein pairs, did not investigate the performance under other cross validations, especially for cross validation on independent lncRNAs and independent proteins. (3) lncRNAs and proteins have abundant biological information, how to select informative features need to further investigate.ResultsUnder a hybrid framework (LPI-HyADBS) integrating feature selection based on AdaBoost, and classification models including deep neural network (DNN), extreme gradient Boost (XGBoost), and SVM with a penalty Coefficient of misclassification (C-SVM), this work focuses on finding new LPIs. First, five datasets are arranged. Each dataset contains lncRNA sequences, protein sequences, and an LPI network. Second, biological features of lncRNAs and proteins are acquired based on Pyfeat. Third, the obtained features of lncRNAs and proteins are selected based on AdaBoost and concatenated to depict each LPI sample. Fourth, DNN, XGBoost, and C-SVM are used to classify lncRNA-protein pairs based on the concatenated features. Finally, a hybrid framework is developed to integrate the classification results from the above three classifiers. LPI-HyADBS is compared to six classical LPI prediction approaches (LPI-SKF, LPI-NRLMF, Capsule-LPI, LPI-CNNCP, LPLNP, and LPBNI) on five datasets under 5-fold cross validations on lncRNAs, proteins, lncRNA-protein pairs, and independent lncRNAs and independent proteins. The results show LPI-HyADBS has the best LPI prediction performance under four different cross validations. In particular, LPI-HyADBS obtains better classification ability than other six approaches under the constructed independent dataset. Case analyses suggest that there is relevance between ZNF667-AS1 and Q15717.ConclusionsIntegrating feature selection approach based on AdaBoost, three classification techniques including DNN, XGBoost, and C-SVM, this work develops a hybrid framework to identify new linkages between lncRNAs and proteins.

Dataset Information

Enhancing data pipelines for forecasting student performance: integrating feature selection with cross-validation.

Supplementary information

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets