IEEE Transactions on Pattern Analysis and Machine Intelligence

An ensemble machine learning approach for determination of the optimum sampling time for evapotranspiration assessment from high-throughput phenotyping data

Applications of Data Mining techniques are increasing on a rapid pace in the area of Higher Education. Faculty Performance Evaluation is a main concern in today’s scenario as the vast amount of data is generated by educational institutes, which contains valuable hidden information. Though, Data Mining techniques have already applied to evaluate the faculty performance but no technique is found adequate in accuracy. In this research, more than one data mining techniques are ensembled in an order for achieving better accuracy rates. Here, we have applied Boosting (Ensemble) Feature Stacking, Random Forest and Generalized Boosted Model which are based on base learners Decision Tree, SVM and Naive Bayes. These ensemble classifiers are applied on Faculty dataset and student results for evaluation of Faculty Performance. Performances of ensemble classifier are evaluated on the basis of Accuracy, Sensitivity and Specificity. We have conducted two studies for this purpose, one with target variable "Pass_Fail" and another with Target variable "Good_Grades". y are used as a measure for evaluating performance.

Evaluating Faculty Performance Using Ensemble Techniques

Machine Learning (ML) methods have been increasingly employed in the genetics domain. ML methods have shown promise in the field of characterizing genetic mutations. Mutations can have significant impact on the activity of the Human Epidermal Growth Factor Receptor (EGFR), a protein instrumental in cell proliferation. Over-activation of EGFR is a major cause of tumor growth. Although many computational methods have been proposed to identify disease causing mutations, these methods are not designed to predict mutation impact on protein activity. We explored feature selection strategies suitable for the small, complex data within this domain and tested a variety of machine learning algorithms. We generated a model achieving 85.9% accuracy and an F-Measure of 0.70 with a Support Vector Machine with a Gaussian radial basis function kernel using a set of 6 features. This classifier combined with others using weighted probability voting achieved an area under the ROC curve of 0.83. INDEX WORDS: Machine Learning, Bioinformatics, Genetics, Mutation, Kinase, Protein, Cancer, EGFR, DNA, Naïve-Bayes, Random Forest, SVM, Nearest Neighbor, Logistic Regression PREDICTION OF CANCER-RELATED MUTATION IMPACT ON PROTEIN ACTIVITY USING MACHINE LEARNING

Prediction of cancer-related mutation impact on protein activity using machine learning

Performance in terms of high accuracy is very much required to any data mining system. A very much influencing factor on the success or failure of the data mining process is the quality of data used in the process. Feature selection is one of the preprocessing techniques in data mining which can prepare a quality data set before feeding it into the data mining process. This paper focuses on the Best First Search technique of feature selection. Feature selection is done using Best First Search. After selecting the features in given data set, it is taken as an input to the data mining process. It is observed that when classification is performed with feature selection, performance of classifier Performance of the classification. The success of a data mining problem heavily depends upon the quality of the data which is the most influencing factor. Moreover, feature selection represents one of the tools which can refine a dataset before presenting it to a learning scheme. In this paper, analyses of a wrapper approach for feature selection, with the purpose of boosting the classification accuracy is done. Experimental evaluations have been performed of feature selection on several data sets. The results showed that feature selection improves the overall performance in classification. Accuracy and speeds up the training process.

Effect of Feature Selection Using Best First Search on the Performance of Classification

Data mining is the practice of analyzing huge quantities of data and shortening it into constructive knowledge. Data Mining is an eternal process which is quite useful in finding understandable patterns and relationships amongst the data. There are various classification techniques available. It is observed that all the techniques don’t work well with all datasets. It is found that when the classifiers are used alone, they are not performing as good as when they are combined using ensembles. Ensemble methods are renowned techniques in order to improve the classification accuracy. Bagging and Boosting are the most common ensemble learning techniques used to improve the classification accuracy. Here, a study on the classification accuracy improvement is carried out in which an experiment is performed using boosting with different datasets from UCI repository.

/pdf/an-experiment-to-improve-classification-accuracy-using-3dn0nm8ehc.pdf

An Experiment to Improve Classification Accuracy Using Ensemble Methods

One of the critical task in data mining is classification. It is very much important in classification to achieve maximum accuracy. In the field of data mining, numerous classifiers are present for the classification task. Each classification techniques have their pros and cons. Some of the techniques work well with certain data sets while other techniques work well with other data sets. There have been many techniques evolved for improving classification accuracy. One of such technique is pre-processing which helps in improving quality of the data. Another method is to combine the classifiers, which will in turn improve the classification accuracy. In this paper, empirical study is been done on various techniques for improving classification accuracy. One of the technique is feature selection, which will select best features from the available features in the data set. Other approach is ensemble learning which combines many classifiers to improve the classification accuracy.

/pdf/improving-classifier-performance-using-feature-selection-397ig4cgd3.pdf

Improving Classifier Performance Using Feature Selection with Ensemble Learning

There are various classifiers available for data classification, selecting the best classifier is one of the critical problems of data classification. Also pre-processing approach to be used is quite important. In this paper, study of various approaches to improve the classification accuracy in data mining is carried out. The purpose of the preprocessing is to gain a high degree of distinct classes before the classifier is trained or tested. Handling noise and outliers is an important aspect in data mining to improve the classification accuracy. High accuracy of classification also depends upon the quality of data being used for classification in data mining. Feature selection is also one of the aspects which can refine the dataset before providing it to the learning algorithm to improve the accuracy of the

A survey on Improving Classification Accuracy in Data Mining

Data Mining is the study to get the knowledge from the huge data sources. It is a technology with huge potential to help the corporate ventures focus on the most important information in their data warehouses or database, so that it will help in making business decisions. Decision making with data mining is very much complex task. Ensemble technique is one of the common strategies to improve the accuracy of classifier. In general ensemble learning is an effective technology that combines the predictions from multiple base classifiers. Most commonly used ensemble techniques are bagging and Boosting. Stacking is also one of the techniques, but it is less widely used. In this paper, we are focusing on bagging technique. An experiment is carried out using bagging with different datasets from UCI repository to study the classification accuracy improvement

Bhavesh Patankar

Papers

Effect of Feature Selection Using Best First Search on the Performance of Classification

An Experiment to Improve Classification Accuracy Using Ensemble Methods

Improving Classifier Performance Using Feature Selection with Ensemble Learning

A survey on Improving Classification Accuracy in Data Mining

Improving Classification Accuracy through ensemble technique in Data Mining