With the help of an electroencephalogram (EEG) the electrical activity of the brain is measured, and this can help identify chronic neurological disorders such as epilepsy. Epileptic episodes are detected by monitoring patients in order to provide preventive measures. Current research studies are using a combination of time and frequency features to recognize epileptic seizures automatically. In order to automatically detect epileptic seizures, different machine learning approaches have been used. Gradient boosting decision tree (GBDT) is a machine learning technique that is known for its efficiency, accuracy, and interpretability. In terms of performance of GBDT, many machine learning tasks such as multi-class classification, learning to rank, etc. have reported competitive performance. In this paper, epileptic seizure recognition data is investigated and split into a binary and multi-class data set for which the GBDT method is applied. In addition, the SHAP (Shapley Additive Explanations) method is used as an explanation tool to interpret the machine learning models that are produced via training for both the binary and the multi-class data set.

This paper presents a parallel mining architecture model intended to be used in blockchain systems to improve transaction speed and network scalability while maintaining decentralization. Typical blockchain validation times are significantly slower than traditional digital transaction systems. The model proposed is intended to allow devices with limited computational power to make meaningful contributions to the blockchain system by introducing parallel proof of work, managed by automated manager nodes. This will allow blockchain systems to be integrated into cloud environments and the internet of things. The proposed model is also intended to address and reduce power consumption problems current blockchain systems face, by allowing the network to validate transactions without the need of high-powered specialty mining machines. Automation and virtualization of network nodes is intended to utilize hardware already online to preform parallel proof of work together in contrast to nodes all competing against each other and wasting electrical power.

In our research, supervised machine learning algorithms were applied to analyze and compare their capability of cancer classification. Our research used eight machine learning algorithms: Decision Tree, Gradient Boosting, K-Nearest Neighbors, Logistic Regression, Naïve Bayes, Neural Network, Random Forest, and Support Vector Machine. Machine learning models were generated by training the algorithms on the TCGA Pancancer HiSeq data set. This data set is an RNA sequencing (RNA-seq) data set consisting of five separate cancer types such as breast invasive carcinoma (BRCA), colon adenocarcinoma (COAD), kidney renal clear cell carcinoma (KIRC), lung adenocarcinoma (LUAD), and Prostate adenocarcinoma (PRAD). The data set was preprocessed with feature selection, oversampling, and normalization techniques. The preprocessed methods were implemented by selecting only the best features and thus removing the inconsequential features, balancing the sample size of each cancer type, and rescaling the values of numeric attributes. Our goal was to determine which algorithm generates a classification model that shows the best performance when categorizing cancer types by employing the following evaluation measures: accuracy, precision, recall, area under curve (AUC) score, F-1 score, and processing time.

Early detection of cancer can significantly increase the chance of successful treatment. This research performs a study on early cancer detection for prostate cancer patients from whom cancer tissue was analyzed with Illumina Hi-Seq ribonucleic acid (RNA) Sequencing (RNA-Seq). Cancer relevant genes with the most significant correlations with the clinical outcome of the sample type (cancer /non-cancer) and the overall survival (OS) were assessed. Traditional cancer diagnosis primarily depends on physicians' experience to identify morphological abnormalities. Gene expression level data can assist physicians in detecting cancer cases at a much earlier stage and thus can significantly improve the potential of patient treatment. In this research, for the classification task, we applied machine learning and data mining approaches to detect cancer versus non-cancer based on gene expression data. Our goal was to detect cancer at the earliest stage. Besides, for the regression task, survival outcomes in prostate cancer patients were performed. Regression trees were built using cancer-sensitive genes along with clinical attribute `Gleason score' as predictors, and the clinical variable `overall survival' as the target variable. Knowledge in the form of rules is one of the vital tasks in data mining as it provides concise statements of easily understandable and potentially valuable information. For the classification model, we derived rules from a decision tree and interpreted these rules for cancer and non-cancer patients. For the regression or survival model, we generated rules for predicting or estimating the survival time of cancer patients. In this study, cancer-relevant genes were analyzed as predictors, although various genes may interact with genes currently known to contribute to cancer. These findings have implications for assessing gene-gene interactions and gene-environment interactions of prostate cancer as well as for other types of cancer.

Breast cancer is the most common cancer in women worldwide. Prevention of breast cancer through risk factors reduction is a significant concern to decrease its impact on the population. Attaining or detecting significant information in the form of rules is the key to prevent breast cancer. Our objective is to find hidden but important knowledge of the form of rules from the risk factors data set of breast cancer. Mining rules is one of the vital tasks of data mining as rules provide concise statement of potentially important information that is easily understood by end users. In this paper, we use association rule mining, a data mining technique to attain information in the form of rules from breast cancer risk factors data that could be useful to initiate prevention strategies. We discovered rules of both breast cancer and non-breast cancer patients so that we can understand and compare the characteristics of both breast cancer and non-breast cancer individuals. The experimental results show that generated or mined rules hold the highest confidence level.

Genetic Programming (GP) is one of the evolutionary computation techniques that is used for the classification process. GP has shown that good accuracy values especially for binary classifications can be achieved, however, for multiclass classification unfortunately GP does not obtain high accuracy results. In this paper, we propose two approaches in order to improve the GP classification task. One approach (GP-K) uses the K-means clustering technique in order to transform the produced value of GP into class labels. The second approach (GP-D) uses a discretization technique to perform the transformation. A comparison of the original GP, GP-K and GP-D was conducted using binary and multiclass datasets. In addition, a comparison with other state-of-the-art classifiers was performed. The results reveal that GP-K shows good improvement in terms of accuracy compared to the original GP, however, it has a slightly longer execution time. GP-D also achieves higher accuracy values than the original GP as well as GP-K, and the comparison with the state-of-the-art classifiers reveal competitive accuracy values.

Service composition is the process of combining services in a specific order to achieve a specific goal, whereby the initial and goal states are determined in advance. The service composition problem is very similar to standard planning problems since the idea is to discover a path between the initial and goal states. In service composition, the composition of services identifies this path. In this paper, we exploit provenance information along with Partially Observable Markov Decision Processes (POMDP) to compose the services automatically. The POMDP method has been used in literature for the purpose of robot planning and navigation. In this research, we argue that due to partial observability of service and system states, the POMDP approach provides better solutions for the QoS-aware service composition in dynamic workflow environments. For the purpose of solving the POMDP, service details and the POMDP distributions are learnt from the provenance store. Provenance data contains information regarding workflows, services, their specifications and execution details. This information facilitates the service composition process to be performed more intelligently and efficiently.