ReliefF Feature Selection Method Research Articles

A micro-XRT image analysis and machine learning methodology for the characterisation of multi-particulate capsule formulations

The application of X-ray microtomography for quantitative structural analysis of pharmaceutical multi-particulate systems was demonstrated for commercial capsules, each containing approximately 300 formulated ibuprofen pellets. The implementation of a marker-supported watershed transformation enabled the reliable segmentation of the pellet population for the 3D analysis of individual pellets. Isolated translation- and rotation-invariant object cross-sections expanded the applicability to additional 2D image analysis techniques. The full structural characterisation gave access to over 200 features quantifying aspects of the pellets' size, shape, porosity, surface and orientation. The extracted features were assessed using a ReliefF feature selection method and a supervised Support Vector Machine learning algorithm to build a model for the detection of broken pellets within each capsule. Data of three features from distinct structure-related categories were used to build classification models with an accuracy of more than 99.55% and a minimum precision of 86.20% validated with a test dataset of 886 pellets. This approach to extract quantitative information on particle quality attributes combined with advanced data analysis strategies has clear potential to directly inform manufacturing processes, accelerating development and optimisation.

A spatially explicit deep learning neural network model for the prediction of landslide susceptibility

With the increasing threat of recurring landslides, susceptibility maps are expected to play a bigger role in promoting our understanding of future landslides and their magnitude. This study describes the development and validation of a spatially explicit deep learning (DL) neural network model for the prediction of landslide susceptibility. A geospatial database was generated based on 217 landslide events from the Muong Lay district (Vietnam), for which a suite of nine landslide conditioning factors was derived. The Relief-F feature selection method was employed to quantify the utility of the conditioning factors for developing the landslide predictive model. Several performance metrics demonstrated that the DL model performed well both in terms of the goodness-of-fit with the training dataset (AUC = 0.90; accuracy = 82%; RMSE = 0.36) and the ability to predict future landslides (AUC = 0.89; accuracy = 82%; RMSE = 0.38). The efficiency of the model was compared to the quadratic discriminant analysis, Fisher's linear discriminant analysis, and multi-layer perceptron neural network. A comparative analysis using the Wilcoxon signed-rank tests revealed a significant improvement of landslide prediction using the spatially explicit DL model over these other models. The insights provided from this study will be valuable for further development of landslide predictive models and spatially explicit assessment of landslide-prone regions around the world.

Artificial Neural Network based Process History Data Model for Gas Turbine Compressor Systems

Gas turbine-based power plants are found to play a vital role in electric power generation and act as spinning reserves for renewable electric power. A robust performance assessment tool is inevitable for a gas turbine system to maintain high operational flexibility, availability, and reliability at different operating conditions. A suitable simulation model of the gas turbine provides detailed information about the system operation under varying ambient and load conditions. This paper illustrates a systematic methodology for process history data-based modelling of a gas turbine compressor system. The ReliefF feature selection method is applied for the proper identification of the parameters influencing the compressor efficiency. Appropriate Artificial Neural Network (ANN) based models are developed for data classification and system modelling of the compressor. The model performance has been validated using actual plant operational data, and the standard deviation of the error in model output was found to be 0.38. A novel approach for suitable integration of data processing methods, machine learning tools and gas turbine domain knowledge has led to the development of a robust compressor model. The model has been utilized for the health assessment of an existing gas turbine compressor, demonstrated through an illustrative case study. The model has been found suitable for parametric analysis of compressor efficiency with operating hours, which is helpful for operational decision-making involving studies on the influence of part-load operation, compressor wash planning, maintenance planning etc.

Cognitive fluorescence sensing to monitor the storage conditions and locate adulterations of extra virgin olive oil

In the present research, storage conditions of extra virgin olive oil (EVOO) have been monitored using cost-effective fluorescence sensors integrated with intelligent algorithms. Three different Spanish pure EVOOs (Arbequina, Cornicabra, and Picual varieties), as well as samples adulterated with expired EVOO, were kept under different environmental conditions: presence or absence of light, bottle material (plastic or glass), and clearness of glass (uncolored or brown glass). To evaluate the effect of these conditions, 54 samples were prepared, and their emission spectra were measured 10 times during a 58-day period. The most statistically relevant information from these spectra was located by the relief-F feature selection method, which led to the design of machine learning-based models. With this aim, up to 158,250 artificial neural network-based models were designed and tested. It was possible to distinguish different storage conditions in terms of light exposure, bottle material, and clearness, as well as locate adulterations. All of the trained and optimized classifiers for all three EVOO varietals reached accuracies ranging between 91 and 100% after meticulous validations. The used technique is fast, portable, user-friendly, inexpensive, and non-destructive, and, therefore, it could be of great use for the olive oil sector as it enables real-time quality control and fraud detection.

Prediction of Concrete Compressive Strength and Slump by Machine Learning Methods

Machine learning methods have been successfully applied to many engineering disciplines. Prediction of the concrete compressive strength (fc) and slump (S) is important in terms of the desirability of concrete and its sustainability. The goals of this study were (i) to determine the most successful normalization technique for the datasets, (ii) to select the prime regression method to predict the fc and S outputs, (iii) to obtain the best subset with the ReliefF feature selection method, and (iv) to compare the regression results for the original and selected subsets. Experimental results demonstrate that the decimal scaling and min‐max normalization techniques are the most successful methods for predicting the compressive strength and slump outputs, respectively. According to the evaluation metrics, such as the correlation coefficient, root mean squared error, and mean absolute error, the fuzzy logic method makes better predictions than any other regression method. Moreover, when the input variable was reduced from seven to four by the ReliefF feature selection method, the predicted accuracy was within the acceptable error rate.

Use of Nonlinear Features for Automated Characterization of Suspicious Ovarian Tumors Using Ultrasound Images in Fuzzy Forest Framework

Ovarian cancer is one of the prime causes of mortality in women. Diagnosis of ovarian cancer using ultrasonography is tedious as ovarian tumors exhibit minute clinical and structural differences between the suspicious and non-suspicious classes. Early prediction of ovarian cancer will reduce its growth rate and may save many lives. Computer-aided diagnosis (CAD) is a noninvasive method for finding ovarian cancer in its early stage which can avoid patient anxiety and unnecessary biopsy. This study investigates the efficacy of a novel CAD tool to characterize suspicious ovarian cancer using Radon-transformed nonlinear features. The obtained dimension of the extracted features is reduced using Relief-F feature selection method. In this study, we have employed the fuzzy forest-based ensemble classifier in contrast to the known crisp rule-based classifiers. The proposed method is evaluated using 469 (non-suspicious: 238, suspicious: 231) subjects and achieved a maximum 80.60 ± 0.5% accuracy, 81.40% sensitivity, 76.30% specificity with fuzzy forest, an ensemble fuzzy classifier using thirty-nine features. The proposed method is robust and reproducible as it uses maximum number subjects (469) as compared to state-of-the-art techniques. Hence, it can be used as an assisting tool by gynecologists during their routine screening.

Wearable Vibration Based Computer Interaction and Communication System for Deaf

In individuals with impaired hearing, determining the direction of sound is a significant problem. The direction of sound was determined in this study, which allowed hearing impaired individuals to perceive where sounds originated. This study also determined whether something was being spoken loudly near the hearing impaired individual. In this manner, it was intended that they should be able to recognize panic conditions more quickly. The developed wearable system has four microphone inlets, two vibration motor outlets, and four Light Emitting Diode (LED) outlets. The vibration of motors placed on the right and left fingertips permits the indication of the direction of sound through specific vibration frequencies. This study applies the ReliefF feature selection method to evaluate every feature in comparison to other features and determine which features are more effective in the classification phase. This study primarily selects the best feature extraction and classification methods. Then, the prototype device has been tested using these selected methods on themselves. ReliefF feature selection methods are used in the studies; the success of K nearest neighborhood (Knn) classification had a 93% success rate and classification with Support Vector Machine (SVM) had a 94% success rate. At close range, SVM and two of the best feature methods were used and returned a 98% success rate. When testing our wearable devices on users in real time, we used a classification technique to detect the direction and our wearable devices responded in 0.68 s; this saves power in comparison to traditional direction detection methods. Meanwhile, if there was an echo in an indoor environment, the success rate increased; the echo canceller was disabled in environments without an echo to save power. We also compared our system with the localization algorithm based on the microphone array; the wearable device that we developed had a high success rate and it produced faster results at lower cost than other methods. This study provides a new idea for the benefit of deaf individuals that is preferable to a computer environment.

Classification of Real Imbalanced Cardiovascular Data Using Feature Selection and Sampling Methods: A Case Study with Neural Networks and Logistic Regression

Coronary Artery Disease (CAD) is very common among the major types of cardiovascular diseases, and there are several studies created with different features including data that is collected from patients for timely diagnosis of CAD.In this study, a dataset with 21 features have been used, and a risk score prediction system has been proposed. The patients were divided into four groups. To determine the effective features of CAD dataset; t-test and Relief-f methods on Logistic Regression Analysis (LRA); Relief-f on Neural Network (NN) feature selection methods were utilized.Sampling methods were used to improve imbalanced form of 4-classed dataset, and the effects of sampling methods were evaluated. Using NN with oversampling and Relief-f feature selection method; the results before the preprocess operations were detected as follows; 72.3% accuracy; after the operations, 84.1% accuracy were achieved with 0.84 sensitivity 0.94 specificity. These statistics obtained from the experiment, by detailed analysis, are the best ones for the CAD data set in this study.Using the feature selection and the sampling methods with the NN substantially improve the prediction accuracy as well as the other metrics. This suggests that these preprocessing methods and the NN may be used together to construct for prediction of the 4-classed imbalanced medical datasets.

Recognition of protozoan parasites from microscopic images: Eimeria species in chickens and rabbits as a case study.

Automated diagnosis and identification of diseases and conditions such as parasites from microscopic images have been mainly carried out by utilizing the object morphological characteristics. The extraction of morphometric features needs the use of highly complex techniques that require computational power. Therefore, in order to reduce this complexity, this paper presents an automated identification based on analyzing three groups of pixel-based feature sets: column features (CF), row features (RF), and the third one (CRF) obtained by merging CF and RF together. For the classification task, K-Nearest Neighbor (KNN) and Artificial Neural Networks (ANN) have been applied. The classification results have been evaluated by adapting a 5-fold cross validation. Additionally, a robust sub-set of the features has been selected by Relieff feature selection method to prevent overfitting, which in turn has improved the final results. Two microscopic image slide databases of a type of protozoan parasites genus called Eimeria in fowls and rabbits have been examined in order to assess the robustness of the proposed methods. The highest accuracy rates obtained when the entire features were used are 85.55% (±0.39%) and 96.6% (±0.82%) from grey-scale level and color images, respectively. These results have been increased by 5% when the feature size is reduced by two thirds when Relieff was utilized. The feature sets have yielded highly accurate results and are expected to make the automatic identification simpler than the analysis of morphological features.

Comprehensive identification of multiple harmonic sources using fuzzy logic and adjusted probabilistic neural network

This paper presents a comprehensive approach based on fuzzy logic and probabilistic neural network (PNN) to identify location, relative level, and type of multiple harmonic sources in power distribution systems. The location and relative level of harmonic sources were determined in the fuzzy stage by interpreting harmonic powers together with network impedances. Then, the type of the harmonic sources was classified in the neural stage using adjusted PNN. In the proposed method, the harmonic powers were considered as classification features. Then, ReliefF feature selection method was used to reduce the redundant data and dimension of features vector. A new modified adaptive imperialist competitive algorithm (MAICA) was proposed to determine the only adjusted parameter of the PNN classifier. Furthermore, a deep belief network (DBN) was applied in the neural stage, and its results were compared with the PNN classifier. The proposed approach was evaluated on IEEE 18-bus and IEEE 69-bus test systems. Unlike the single point methods, the presented method provides information on multiple harmonic sources in the whole of the distribution system. The results show that the comprehensive approach identifies the multiple harmonic sources with high accuracy.

Recognition and assessment of different factors which affect flicker in wind turbines

Similar to other distributed generation sources, wind turbines cause power quality disturbances (PQDs) issues in power systems. One of the most important PQDs that has bad effects on sensitive loads is flicker. In this study, an algorithm is presented for assessment and recognition of different factors causing flicker of wind turbines. Some aerodynamic factors (wind shear and tower shadow) and some mechanical factors (blade pitching errors, gearbox tooth crash and turbine blade break down) are modelled using fixed speed wind turbines. Then, wavelet transform and S-transform are used to extract some dominant features voltage. Then, in order to avoid large dimension of feature vector, Relieff feature selection method is applied to extracted features. The probabilistic neural network (PNN) is used to classify above-mentioned factors. The only adjusted parameter of the PNN classifier is determined by using the particle swarm optimisation technique. Moreover, the short-term severity of flicker (Pst) is calculated for each type of fault as extra features to increase the severability of extracted features. Results show that the classifier can detect different causes of flicker event with high detection accuracy.

Forecasting of Wind Speed using Feature Selection and Neural Networks

Wind energy is rapidly increasing and it is becoming a significant contributor to the electricity grid. Wind speed is an important factor in wind power production and integration. This paper presents a wind speed forecasting using feature selection method and bagging neural network. Feature selection plays an essential role in the machine learning environment and especially in the prediction task. The ReliefF feature selection method is used for identification of necessary features for wind speed forecast and reduces the complexity of the model. A detailed investigation is to forecast wind speed with meteorological time series data as input variable using a bagging neural network. Performance is evaluated in terms of mean square error when using the feature selection method with the bagging neural network.

Ensemble Learning in Data Mining of Fetal Cardiotocograms

ReliefF feature selection and LogitBoost ensemble learning method are employed in the data mining procedure of 2126 fetal cardiotocograms (CTGs). Based on 10 critical features selected by ReliefF and the full 21 features, LogitBoost algorithm almost outperforms the other three ensemble learning methods of Stacking, Bagging and AdaBoostM1 in ACC (%) and AUC in classification, and the ACC (%) and AUC of LogitBoost algorithm are achieved to 94.45% and 0.977 based on the critical features from ReliefF.

Analyzing Online Review Helpfulness Using a Regressional ReliefF-Enhanced Text Mining Method

Within the emerging context of Web 2.0 social media, online customer reviews are playing an increasingly important role in disseminating information, facilitating trust, and promoting commerce in the e-marketplace. The sheer volume of customer reviews on the web produces information overload for readers. Developing a system that can automatically identify the most helpful reviews would be valuable to businesses that are interested in gathering informative and meaningful customer feedback. Because the target variable---review helpfulness---is continuous, common feature selection techniques from text classification cannot be applied. In this article, we propose and investigate a text mining model, enhanced using the Regressional ReliefF (RReliefF) feature selection method, for predicting the helpfulness of online reviews from Amazon.com. We find that RReliefF significantly outperforms two popular dimension reduction methods. This study is the first to investigate and compare different dimension reduction techniques in the context of applying text regression for predicting online review helpfulness. Another contribution is that our analysis of the keywords selected by RReliefF reveals meaningful feature groupings.