Supervised Learning Classification Method Research Articles

Learning matrix profile method for discord-based attribution of electricity consumption pattern behavior

This paper frames itself in analyzing complex electricity load measurements to attribute the consumption behavior to a specific entity. Attribution may be seen as the first step in several grid-based activities, like energy management, privacy, and identification of illicit activities. This work proposes and tests a novel approach that utilizes consumption discords as the analysis carrier for attributing multiple load patterns to specific consumers. The discord-driven analysis is performed utilizing the synergism of the Matrix Profile with each of two well-established supervised learning classification methods: the K-Nearest-Neighbor and Support-Vector-Machine. The proposed approach is applied in attributing electricity consumption pattern behaviors to a set of academic institutions that are comprised of multiple academic units. Notably, multiple units within the same entity exhibit different consumption behavior, thus imposing a high challenge in attributing the unit’s consumption behavior to its entity of origin when there is a variety of targeted entities. Obtained results demonstrate that the load discord-based MP-KNN and MP-SVM combinations provide higher identification accuracy than the state-of-the-art method of performing supervised classification with a feed-forward artificial neural network of the full load patterns. The identification accuracy of the proposed method outperformed the ANN load classification by approximately 13%.

Data reduction based on machine learning algorithms for fog computing in IoT smart agriculture

Smart agriculture applications that analyse and manage agricultural yield using IoT systems may suffer from intermittent operation due to cloud disconnections commonly occurring in rural areas. A fog computing solution enables the IoT system to process data faster and deal with intermittent connectivity. However, the fog needs to send a high volume of data to the cloud and this can cause link congestion with unusable data traffic. Here we propose an approach to collect and store data in a fog-based smart agriculture environment and different data reduction methods. Sixteen techniques for data reduction are investigated; eight machine learning (ML) methods combined with run-length encoding, and eight combined with Huffman encoding. Our experiment uses two real data sets, where the first contains air temperature and humidity values, and the second has soil moisture and temperature conditions. The fog filters cluster the unlabelled data using unsupervised machine learning algorithms that group data into categories according to their value ranges in all experiments. Supervised learning classification methods are also used to predict the class of data samples from these categories. After that, the fog filter compresses the identified categories using two data compression techniques, run-length encoding (RLE) and the Huffman encoding, preserving the data time series nature. Our results reveal that a k-means combined with RLE method achieved the highest reduction, where the fog needed to store and transmit only 3%–6% of the original data generated by sensors. • It is a challenge to manage a massive amount of data generated by sensors in IoT. • Combining machine learning with data compression results in a larger data reduction. • Depending on the context, the fog needs to decide which classifier should use.

Continual learning classification method with the weighted k-nearest neighbor rule for time-varying data space based on the artificial immune system

Supervised learning classification methods play an important role in various fields. However, most of them cannot effectively classify the data from time-varying data spaces, for they lack continual learning ability. Inspired by the intelligent mechanism that memory cells of the biological immune system can evolve with the evolution of invaders, a continual learning classification method with the weighted k-nearest neighbor rule for time-varying data space (WKNN-CLCMTVD) is proposed. Memory cells selected by the weighted k-nearest neighbor rule are used to identify the type of testing data. Memory cells are continuously cultivated, updated, and eliminated through learning testing data to improve the classification ability of WKNN-CLCMTVD. It degenerates into a standard supervised learning classification method when all data independent of time. Take experiments on twenty benchmark datasets, a 2-dimensional synthetic dataset, and XJTU-SY rolling element bearing accelerated life test datasets to evaluate the performance of WKNN-CLCMTVD. Results show that it has better classification ability for time-invariant data and outperforms the other methods for time-varying data space. Results also show that its running speed is far faster than that of other continual learning classification methods.

Comparison of Distance Models on K-Nearest Neighbor Algorithm in Stroke Disease Detection

Stroke is a cardiovascular (CVD) disease caused by the failure of brain cells to get oxygen supply to pose a risk of ischemic damage and result in death. This Disease can detect based on the similarity of symptoms experienced by the sufferer so that early steps can be taking with appropriate counseling and treatment. Stroke detecting requires a machine learning method. In this research, the author used one of the supervised learning classification methods, namely K-Nearest Neighbor (K-NN). K-NN is a classification method based on calculating the distance to training data. This research compares the Euclidean, Minkowski, Manhattan, Chebyshev distance models to obtain optimal results. The distance models have been tested using the stroke dataset sourced from the Kaggle repository. Based on the test results, the Chebyshev model has the highest levels of accuracy compared to the other three distance models with an average accuracy value of 95.49%, the highest accuracy of 96.03%, at K = 10. The Euclidean and Minkowski distance models have the same level of accuracy at each K value with an average accuracy value of 95.45%, the highest accuracy of 95.93% at K = 10. Meanwhile, Manhattan has the lowest average compared to the other distance models, which is 95.42% but has the highest accuracy of 96.03% at the value of K = 6

Continual learning classification method for time-varying data space based on artificial immune system

Classification methods play an important role in many fields. However, they cannot effectively classify the samples from sample spaces that are varying with time, for they lack continual learning ability. A continual learning classification method for time-varying data space based on artificial immune system, CLCMTVD, is proposed. It is inspired by the intelligent mechanism that memory cells of the biological immune system can recognize and eliminate previous invaders when they attack again very fast and more efficiently, and these memory cells can evolve with the evolution of previous invaders. Memory cells were continuously updated by learning testing data during the testing stage, thus realize the self-improvement of classification performance. CLCMTVD changes a linearly inseparable spatial problem into many classification problems of several different times, and it degenerates into a common supervised learning classification method when all data independent of time. To assess the performance and possible advantages of CLCMTVD, the experiments on well-known datasets from UCI repository, synthetic data and XJTU-SY rolling element bearing accelerated life test datasets were performed. Results show that CLCMTVD has better classification performance for time-invariant data, and outperforms the other methods for time-varying data space.

Continual learning classification method with constant-sized memory cells based on the artificial immune system

Most classification methods cannot further improve their classification performance by learning the testing data during the testing stage, for lacking continual learning ability. A new classification method, continual learning classification method with constant-sized memory cells based on the artificial immune system (C-CLCM), is proposed. It is inspired by the continual learning mechanism of the biological immune system. C-CLCM gradually enhances its classification performance by continually learning the testing data especially the new types of labeled data and new types of unlabeled data during the testing stage. At the same moment, it updates the existing memory cells and culture new types of memory cells. C-CLCM degenerates into a common supervised learning classification method under certain conditions. To assess its performance and possible advantages, the experiments on well-known datasets from the UCI repository were performed. Results show that C-CLCM has better classification performance when it degenerates into a common supervised learning classification method. It outperforms the other methods when the training data do not cover all types. The less type of training, the more advantages it has.

Sentiment Analysis on KAI Twitter Post Using Multiclass Support Vector Machine (SVM)

Information in form of unstructured texts is increasing and becoming commonplace for its existence on the internet. This information is easily found and utilized by business people or companies through social media. One of them is Twitter. Twitter is ranked 6th as a social media that is widely accessed today. The use of Twitter has the disadvantage of unstructured and large data. Consequently, it is difficult for business people or companies to know opinion towards service with limited resources. To Make it easier for businesses know the public's sentiment for better service in the future, public sentiment on Twitter needs to be classified as positive, neutral, and negative. The Multiclass Support Vector Machine (SVM) method is a supervised learning classification method that handles three classes classification. This paper uses One Against All (OAA) approach as a method to determine the class. This paper contains the results of classifying OAA Multiclass SVM methods with five different weighting features unigram, bigram, trigram, unigram+ bigram, and word cloud for analyzing tweet data, finding the best accuracy and important feature when processed with large data. The highest accuracy is the unigram TF-IDF model combined with the OAA Multiclass SVM with gamma 0.7 is 80.59.

Condition Assessment of Civil Structures for Structural Health Monitoring Using Supervised Learning Classification Methods

Structural health monitoring is an essential process for ensuring the safety and serviceability of civil structures. When a structure suffers from damage, it is necessary to implement maintenance programs for returning the structural performance and integrity to its initial normal condition. An important challenge is that the structure of interest may be damaged even after a sophisticated maintenance program. This conveys the great necessity of performing the second level of structural condition assessment and damage detection of maintained structures. To achieve this aim, this article proposes a novel methodology using the concept of supervised learning. The main objective of the proposed methodology is to train various supervised learning classifiers using a training dataset that consists of features regarding both the undamaged and damaged states of the structure before the maintenance program in the first level. Once the classifiers have been trained, one attempts to predict the class labels of test samples associated with the current state of the structure after the maintenance program during the second level. According to the predefined class labels of the training and test samples in the first stage, it is feasible to recognize the current state of the maintained structure in the second level and detect potential damage. The major contribution of this article is to introduce the concept of supervised learning for damage detection in an innovative manner. A numerical concrete beam and an experimental laboratory frame are used to demonstrate the effectiveness and applicability of the proposed methodology. Results show that this methodology is a practical and reliable tool for structural condition assessment and damage detection of maintained structures.

Deteksi Dini Status Keanggotaan Industri Kebugaran Menggunakan Pendekatan Supervised Learning

In the fitness industry, the number of members is a major factor for the sustainability of its business. The ability of managers and trainers to detect members who represent traits to quit membership is critical. Four supervised learning classification methods like Support Vector Machine, Random Forest, K-Nearest Neighbor, and Artificial Neural Network were used to generate early detection using two variants of datasets that have different amounts of data. Classification results are separated into three different zones, which are Green Zone, Yellow Zone, and Red Zone. Artificial Neural Network methods using backpropagation training give 99.90% of accuracy on a dataset which has more amount of data. The evaluation has been done using the confusion matrix and AUC-ROC curves.

Continual learning classification method with new labeled data based on the artificial immune system

In this paper, a new supervised learning classification method, continual learning classification method with new labeled data based on the artificial immune system (CLCMNLD), is proposed as a new way to improve the classification performance in real-time by continually learning the new labeled data during the testing stage. It is inspired by the mechanism that vaccines can enhance immunity. New types of memory cells were continuously cultured by learning new labeled data during the testing stage. CLCMNLD will degenerate into a common supervised learning classification method when there is no new labeled data comes out during the testing stage. The effectiveness of the proposed CLCMNLD is tested on twenty well-known datasets from the UCI Machine Learning Repository that are commonly used in the domain of data classification. The experiments reveal that CLCMNLD has better classification performance when it degenerates into a common supervised learning classification method, and it outperforms the other methods when there are some new labeled data comes out during the testing stage. The more types of new labeled data, the more advantages it has.

Implementasi Sistem Crowdsourced Labelling Berbasis Web dengan Metode Weighted Majority Voting

Crowdsourced Labelling is a large scale data labelling process, solicits a large group of people to label the data, usually via Internet. This paper discusses about design and implementation of Web-based Crowdsourced Labelling. Supervised learning classification methods need labelled training data for its training phase. Unfortunately, in many cases, there aren’t any already available labelled training data. Large scale data labelling is a tedious and time consuming work. This research develops a web-based crowdsourced labelling which able to solicit a large group of people as data labeler to speed up the data labelling process. This system also allows multiple labeler for every data. The final label is calculated using Weighted Majority Voting method. We grabbed and used Facebook comments from the two candidates’ Facebook Page of 2014 Indonesian Presidential Election as testing data. Based on the testing conducted we can conclude that this system is able to handle all the labelling steps well and able to handle collision occurred when multiple labeler labelling a same data in the same time. The system successfully produces final label in CSV format, which can be processed further with many sentiment analysis tools or machine learning tools.
 Index Terms - Crowdsources labeling, web-based system, supervised learning, weighted majority voting.

Entropy analysis to classify unknown packing algorithms for malware detection

The proportion of packed malware has been growing rapidly and now comprises more than 80 % of all existing malware. In this paper, we propose a method for classifying the packing algorithms of given unknown packed executables, regardless of whether they are malware or benign programs. First, we scale the entropy values of a given executable and convert the entropy values of a particular location of memory into symbolic representations. Our proposed method uses symbolic aggregate approximation (SAX), which is known to be effective for large data conversions. Second, we classify the distribution of symbols using supervised learning classification methods, i.e., naive Bayes and support vector machines for detecting packing algorithms. The results of our experiments involving a collection of 324 packed benign programs and 326 packed malware programs with 19 packing algorithms demonstrate that our method can identify packing algorithms of given executables with a high accuracy of 95.35 %, a recall of 95.83 %, and a precision of 94.13 %. We propose four similarity measurements for detecting packing algorithms based on SAX representations of the entropy values and an incremental aggregate analysis. Among these four metrics, the fidelity similarity measurement demonstrates the best matching result, i.e., a rate of accuracy ranging from 95.0 to 99.9 %, which is from 2 to 13 higher than that of the other three metrics. Our study confirms that packing algorithms can be identified through an entropy analysis based on a measure of the uncertainty of the running processes and without prior knowledge of the executables.

Filtering big data from social media – Building an early warning system for adverse drug reactions

ObjectivesAdverse drug reactions (ADRs) are believed to be a leading cause of death in the world. Pharmacovigilance systems are aimed at early detection of ADRs. With the popularity of social media, Web forums and discussion boards become important sources of data for consumers to share their drug use experience, as a result may provide useful information on drugs and their adverse reactions. In this study, we propose an automated ADR related posts filtering mechanism using text classification methods. In real-life settings, ADR related messages are highly distributed in social media, while non-ADR related messages are unspecific and topically diverse. It is expensive to manually label a large amount of ADR related messages (positive examples) and non-ADR related messages (negative examples) to train classification systems. To mitigate this challenge, we examine the use of a partially supervised learning classification method to automate the process. MethodsWe propose a novel pharmacovigilance system leveraging a Latent Dirichlet Allocation modeling module and a partially supervised classification approach. We select drugs with more than 500 threads of discussion, and collect all the original posts and comments of these drugs using an automatic Web spidering program as the text corpus. Various classifiers were trained by varying the number of positive examples and the number of topics. The trained classifiers were applied to 3000 posts published over 60days. Top-ranked posts from each classifier were pooled and the resulting set of 300 posts was reviewed by a domain expert to evaluate the classifiers. ResultsCompare to the alternative approaches using supervised learning methods and three general purpose partially supervised learning methods, our approach performs significantly better in terms of precision, recall, and the F measure (the harmonic mean of precision and recall), based on a computational experiment using online discussion threads from Medhelp. ConclusionsOur design provides satisfactory performance in identifying ADR related posts for post-marketing drug surveillance. The overall design of our system also points out a potentially fruitful direction for building other early warning systems that need to filter big data from social media networks.

Cell-Based Fuzzy Metrics Enhance High-Content Screening (HCS) Assay Robustness

High-content screening (HCS) allows the exploration of complex cellular phenotypes by automated microscopy and is increasingly being adopted for small interfering RNA genomic screening and phenotypic drug discovery. We introduce a series of cell-based evaluation metrics that have been implemented and validated in a mono-parametric HCS for regulators of the membrane trafficking protein caveolin 1 (CAV1) and have also proved useful for the development of a multiparametric phenotypic HCS for regulators of cytoskeletal reorganization. Imaging metrics evaluate imaging quality such as staining and focus, whereas cell biology metrics are fuzzy logic–based evaluators describing complex biological parameters such as sparseness, confluency, and spreading. The evaluation metrics were implemented in a data-mining pipeline, which first filters out cells that do not pass a quality criterion based on imaging metrics and then uses cell biology metrics to stratify cell samples to allow further analysis of homogeneous cell populations. Use of these metrics significantly improved the robustness of the monoparametric assay tested, as revealed by an increase in Z′ factor, Kolmogorov-Smirnov distance, and strict standard mean difference. Cell biology evaluation metrics were also implemented in a novel supervised learning classification method that combines them with phenotypic features in a statistical model that exceeded conventional classification methods, thus improving multiparametric phenotypic assay sensitivity.

EOG-based drowsiness detection: Comparison between a fuzzy system and two supervised learning classifiers

AbstractDrowsiness is a serious problem, which causes a large number of car crashes every year. This paper presents an original drowsiness detection method based on the fuzzy merging of several eye blinking features extracted from an electrooculogram (EOG). These features are computed each second using a sliding window. This method is compared to two supervised learning classifiers: a prototype nearest neighbours and a multilayer perceptron. The comparison has been carried out on a substantial database containing 60 hours of driving data from 20 different drivers. The method proposed reaches very good performances with 82% of true detections and 13% of false alarms on 20 different drivers without tuning any parameters. The best results obtained by the supervised learning classification methods are only 72% of true detection and 26% of false alarms, which is far worse than the fuzzy method. It is shown that the fuzzy method overtakes the other methods because it is able to take into account the fact that drowsiness symptoms occur simultaneously and in a repetitive way on the different features during the epoch to classify, which is of importance in the drowsiness decision-making process.