Performance Analysis Of Classifiers Research Articles

Comprehensive Performance Analysis of Classifiers in Diagnosis of Epilepsy

Epilepsy becomes one of the most frequently arising brain disorder, and it is marked by the unexpected occurrence of frequent seizures. In this study, the University of the Boon Database with ictal seizure disorder diagnosis of the epilepsy is classified by making use of the expectation maximization features as dimensionality reduction technique followed by the nonlinear model, namely, Gaussian mixture model, logistic regression, firefly algorithm, and hybrid model such as cuckoo search with Gaussian mixture model and firefly algorithm with the Gaussian mixture model which are the classifiers used for the diagnosis of epilepsy from the electroencephalogram signals. The performance of the classifiers is analyzed based on performance index, sensitivity, specificity, accuracy, mean square error, good detection rate, and error rate. The most promising outcome in this work indicates expectation maximization features are applied as the dimensionality reduction technique and the hybrid model Cuckoo search with the Gaussian mixture model outperforms with classification accuracy of 92.19%, performance index of 81.43%, good detection rate of 83.48%, and with low error rate of 15.62%, among other classifiers.

Performance Analysis of classifiers in Detection of Abnormalities from Ultrasonic Images-A Review

In this paper the ultrasonic abdominal diseases are classified using various classifiers and their performances are analysed. Early diagnosis of abdominal disease is very important in deciding the proper treatment process. To carry out any research work without the comparison of the proposed one with the other already existing method is not effective to give better result. Objective of this paper is to review the classification methods based on the standard parameters like, Sensitivity, Specificity, Accuracy. In this review four classification algorithms Naive Bays Classifier, Support Vector Machine (SVM), k-Nearest Neighbor (kNN), Artificial Neural Networks(ANN) are compared.The SVM classifier attained higher accuracy of 98.33 % when compared with all the classifiers.

Performance analysis of classifiers for colon cancer detection from dimensionality reduced microarray gene data

AbstractCancer disease is accountable for many deaths that are over 9.6 million in 2018 and roughly one out of six deaths occur because of cancer worldwide. The colon cancer is the second prominent source of death of around 1.8 million cases. This research is inclined to detect the colon cancer from microarray dataset. It will aids the experts to distinguish the cancer cells from normal cells for appropriate determination and treatment of cancer at earlier stages that leads to increase the survival rate of the patients. The high dimensionality in microarray dataset with less samples and more attributes creates lag in the detection capability of the classifier. Hence there is a need for dimensionality reduction techniques to preserve the significant genes that are prominent in the disease classification. In this article, at first ANOVA method used to select the best genes and then principal component analysis (PCA) and fuzzy C‐means clustering (FCM) techniques are further employed to choose relevant genes. The PCA and FCM features are classified using model, discriminant, regression, hybrid, and heuristic‐based classifiers. The attained results show that the heuristic classifier with PCA features is encapsulated an average classification accuracy of 97.92% for classifying both the colon cancer and normal samples. Also, for FCM features, the Heuristic classifier is maintained at an average classification accuracy of 99.48% and 97.92% for classifying the colon cancer and normal samples, respectively. The Heuristic classifier outperforms with high accuracy than all other classifiers in the classification of colon cancer.

Performance Analysis of Classifiers in Identifying NREM Sleep and Awaken Stages from EEG Signals

Electroencephalogram is a medical procedure which helps in analyzing the activities of the brain through electrical signals. In this paper a simple classification technique of EEG signal into two stages as NREM sleep and awaken stages had been undertaken. Classifying these stages helps the physician to understand the patient's sleep disorder by knowing whether the person's brain is in NREM sleep or awaken stages. Physionet EEG signals are samples of 256 signals per second for 10 seconds duration is used in this work. Then the EEG samples properties are analyzed through various parameters like statistical features, entropy Pearson correlation coefficient, Power spectral density, scatter plots and Hilbert transform plots. The classification of NREM sleep and awaken stage is performed by the ten different classifiers broadly grouped into non linear and hybrid one. The classifiers used include Linear Regression, Non Linear Regression, Logistic Regression, Principal Component Analysis, Kernel Principal Component Analysis, Expectation Maximization, Compensatory Expectation Maximization, Expectation Maximization with Logistic Regression Compensatory Expectation Maximization with Logistic Regression, and Firefly. The performances of the classifiers are analyzed using regular parameters like sensitivity, accuracy, specificity, performance index. The highest accuracy of 95.575% is achieved with linear regression for awaken signal and an accuracy of 95.315% is achieved using kernel PCA for sleep signal.

Efficient Role of Machine Learning Classifiers in the Prediction and Detection of Breast Cancer

The major disease in society is Breast Cancer in women worldwide and 27% of women are affected in cancer. Machine learning classifier is suitable for the physicians to make perfect diagnosis with a low cost and time. Comparative performance analysis of classifier in needed to achieve accurate diagnosis as the medical data consist of high dimension data which are noisy in nature. In this study different machine learning techniques of classifiers are applied to Breast cancer datasets. The cancer rates increase in India that increase in the early thirties but its reach at a peak point at ages 50-64 years. As per NICPR report among 28 women there is a woman who affected breast cancer disease. But this ration will change in urban population, one woman is affected among 22 women. Where in rural this ration is increase as 1 in 60 women? The cure rate of the patient may improve by early diagnosis and treatment which may increase prolonging their lives. Here we have developed a model to identify a cancer cell is benign or malignant; we have used machine learning technique classifiers. Here we have to identify a suitable technique for predicting the disease under different operational conditions and datasets. The result analysis shows that SVM finds out to be a suitable option for identifying different performance matrices such as Sensitivity, Accuracy, error, and Specificity.

Interrogation of Sentiment Perusing With Hash Counting Vectorizer and Term Inverse Frequency Transformer Using Machine Learning Classification

With the fast growing technology, the business is moving towards increasing their profit by interpreting the customer satisfaction. The customer satisfaction can be analyzed in many ways. It is the responsibility of the business to analyze the customer satisfaction in order to improve their turnover and profit. With the current trend, the customers are giving their feedback through mobile and internet. With this overview, this paper attempts to analyze the sentiment of the customer feedback for the movie. The sentiment Analysis on movie Review dataset from the KAGGLE Machine learning repository is used for implementation. The type of sentiment classes is predicted through the following ways. Firstly, the sentiment count for each class is displayed and the top feature words for each sentiment class are also extracted from the dataset. Secondly, the dataset is sampled with counting vectorizer and then fitted with the classifiers like Logistic Regression Classifier, Linear SVM Classifier, Multinomial Naives Bayes Classifier, Gradient Boosting Classifer, Guassian Naive Bayes Classifier, Random Forest Classifier, Decision Tree Classifier and and Extra Tree Classifier. Thirdly, the dataset is sampled with Hashing vectorizer and then fitted with the above specified classifiers. Fourth, the dataset is sampled with TFIFD vectorizer and then fitted with the above specified classifiers. Fifth, the dataset is sampled with TFIFD Transformer and then fitted with the above specified classifiers. Sixth, the Performance analysis of classifiers is performed by analyzing the metrics like Precision, Recall, Fscore and Accuracy. The implementation is carried out using python code in Spyder Anaconda Navigator IP Console. Experimental results shows that the analysis of sentiment done by the random forest classifier is found to be more effective with the Accuracy of 89% for Counting vectorizer and TFIFD transformer, Accuracy of 87% for Hashing vectorizer and Accuracy of 88% for TFIFD vectorizer.

Performance Analysis of Classifiers in Identification of Dry and Wet Spells During the Monsoon Period

This paper aims to identify the rainy and non rainy period of the Indian Peninsular monsoon. The IMD observations were taken over a period of one hundred days from 23rd October 2018 to 30th January 2019 with ninety five wet days and five dry dates. Ten observational features like Max, Min and Average temperatures, Rain fall wind Speed, atmospheric Pressure, Illumination, Visibility, relative cloud density and relative humidity are acquired from the IMD data for peninsular India. These features are further reduced by four statistical parameters such as, mean, variance, skewness and kurtosis. Histogram plots show that the measured features and their statistical parameters follows a non linear pattern. Therefore, a group of five classifiers namely, non linear regression, linear regression, Expectation Maximization, logistic regression and Bayesian linear Discriminant are used to analyze the classification efficiency. All the classifiers attains more the 85% of Classification accuracy (average) in the both dry and wet spell period of monsoon observation

Multi-Agent-Based Performance Analysis of Classifiers for Breast Tumours

The challenging effect of selecting the best classifier among many classifier algorithms has been a big problem in data mining. Machine learning is widely used in bioinformatics and particularly in breast cancer diagnosis. This study is based on developing and evaluating different classifier algorithm (k-NN, J48, Decision table, Decision stump, and Naive Bayes) in order to find the best among them using multi-agent platform and MYSQL for the diagnosis of breast tumors based on associated symptoms and risk factors of cancer diseases. Java Agent Development Environment (JADE) was used for the modeling and simulation. The results and the accuracy score were tested with a breast tumor clinical datasets which were gotten and formed from FMC Yola and FMC Gombe in Nigeria using 10- fold Cross-validation method. The results of the analysis reveal that k-NN classifier has a greater performance capability over other classification algorithms; hence, it is selected to be the best among the tested classifiers with higher accuracy score and lower false positive rate value.

Performance analysis of biometric recognition system based on palmprint

Nowadays biometric authentication is one of the most preferred choice and in that also palmprint is becoming the most widely accepted technique because it can be captured easily and the algorithms can be implemented with ease. Detection of coarse lines in palm images is easy to the point that these lines can be revealed even by using a low determination camera. The integration of palmprint recognition with some other biometric recognition system does not require any special capture devices, so it can be done easily. Hence, this method suites best for individual verification. This research work is related to palmprint recognition in which the main aim is related to software based performance analysis of classifiers. The entire work is divided into three major phases namely palmprint pre-processing, feature extraction and accuracy estimation. Different functions and operators related to binarization, morphology and other operations are used for pre-processing; similarly feature extraction is carried out by using Frangi filter, Freak descriptor and FAST algorithm. After the final feature extraction procedure, classifiers are used to predict the accuracy in identifying right and wrong features for a given subject. This research work uses two classifiers namely discriminant analysis and K-nearest neighbor algorithm (KNN) where discriminant analysis shows 77% accuracy and KNN gives a better accuracy of 93%.

Performance analysis of classifiers for EMG signal of different hand movements

In the design of a health monitoring system, Electromyography (EMG) signal is one of the key parameters. So it is very important to utilise the EMG signal carefully. In this paper, different classification methods have been used to classify the EMG signals. EMG signals have been extracted from five different subjects corresponding to their eight different motions of the right hand using LabVIEW. The classification techniques used includes k-NN, naive Bayes and Artificial Neural Network (ANN) classifiers. Five feature vectors used are mean absolute value, average band power, standard deviation, peak to peak root mean square value and root mean square value to learn the classifier. From the results obtained, it has been observed that the performance of ANN classifier in terms of classification accuracy and time required to classify is the best among the three classifiers considered for EMG signal analysis. ANN has 100% classification efficiency for classification of EMG signals obtained from different subjects relative to their hand motion. Based upon better classification efficiency, a better health monitoring system can be manufactured.

Performance analysis of classifier for brain tumor detection and diagnosis

Indefinite and uncontrollable growth of cells leads to tumors in the brain. The early diagnosis and proper treatment of brain tumors are essential to prevent permanent damage to the brain or even patient death. Accurate data regarding the position of the tumor and its size are essential for effective treatment. Hence, an entirely computerized automatic system to provide accurate tumor data is compulsory for physicians. Such developments are necessary to diagnose brain tumors during brain surgery. Brain magnetic resonance (MR) images are proposed for the detection and segmentation of the tumor region via a completely automatic and highly accurate method. The approach discussed in this paper employs an adaptive neuro fuzzy inference system (ANFIS) based on the automatic seed point selection range. The pixels intensity of the proposed algorithm is not dependent on the tumor type. The tumor’s segmentation results are evaluated based on various criteria, including similarity index (SI), overlap fraction (OF), extra fraction (EF) and positive predictive value (PPV), which corresponded to values of 0.817%, 0.817%, 0.182%, and 0.817%, respectively, in this study. These results indicate that the approach proposed in this study performs better compared to many conventional processes. The significance of this work is the differentiation of brain abnormalities from the healthy brain tissue.