Random Forest Bagging and X-Means Clustered Antipattern Detection from SQL Query Log for Accessing Secure Mobile Data

In the current ongoing crisis, people mostly rely on mobile phones for all the activities, but query analysis and mobile data security are major issues. Several research works have been made on efficient detection of antipatterns for minimizing the complexity of query analysis. However, more focus needs to be given to the accuracy aspect. In addition, for grouping similar antipatterns, a clustering process was performed to eradicate the design errors. To address the above-said issues and further enhance the antipattern detection accuracy with minimum time and false positive rate, in this work, Random Forest Bagging X-means SQL Query Clustering (RFBXSQLQC) technique is proposed. Different patterns or queries are initially gathered from the input SQL query log, and bootstrap samples are created. Then, for each pattern, various weak clusters are constructed via X-means clustering and are utilized as the weak learner (clusters). During this process, the input patterns are categorized into different clusters. Using the Bayesian information criterion, the similarity measure is employed to evaluate the similarity between the patterns and cluster weight. Based on the similarity value, patterns are assigned to either relevant or irrelevant groups. The weak learner results are aggregated to form strong clusters, and, with the aid of voting, a majority vote is considered for designing strong clusters with minimum time. Experiments are conducted to evaluate the performance of the RFBXSQLQC technique using the IIT Bombay dataset using the metrics like antipattern detection accuracy, time complexity, false-positive rate, and computational overhead with respect to the differing number of queries. The results revealed that the RFBXSQLQC technique outperforms the existing algorithms by 19% with pattern detection accuracy, 34% minimized time complexity, 64% false-positive rate, and 31% in terms of computational overhead.

Knowledge-based Residual Learning

Small data has been a barrier for many machine learning tasks, especially when applied in scientific domains. Fortunately, we can utilize domain knowledge to make up the lack of data. Hence, in this paper, we propose a hybrid model KRL that treats domain knowledge model as a weak learner and uses another neural net model to boost it. We prove that KRL is guaranteed to improve over pure domain knowledge model and pure neural net model under certain loss functions. Extensive experiments have shown the superior performance of KRL over baselines. In addition, several case studies have explained how the domain knowledge can assist the prediction.

Effects of Parameter Change on Neural Network and Deep Neural Network

Parameters in neural network such as number of neurons and layers have a direct effect on efficiency and accuracy of the network. On the other hand, it is observed that structure of a network is dependent on data complexity. It is perceived that choosing a complex network for a simple data (e.g. linearly separable data) may not only result in a weak learner but also an increase in computation time. Here, accuracy of the network has been challenged by adding Gaussian Noise to these data to investigate the effect of noise. This additive noise flags overcomplexity thus, decision boundary is generalized which it directs the learner to have better classification as if the number of parameters have been decreased. Throughout these comparisons, other constrains have been minimized in MATLAB to solely observe the effect of each change.

Effects of Parameter Change on Neural Network and Deep Neural Network

Parameters in neural network such as number of neurons and layers have a direct effect on efficiency and accuracy of the network. On the other hand, it is observed that structure of a network is dependent on data complexity. It is perceived that choosing a complex network for a simple data (e.g. linearly separable data) may not only result in a weak learner but also an increase in computation time. Here, accuracy of the network has been challenged by adding Gaussian Noise to these data to investigate the effect of noise. This additive noise flags overcomplexity thus, decision boundary is generalized which it directs the learner to have better classification as if the number of parameters have been decreased. Throughout these comparisons, other constrains have been minimized in MATLAB to solely observe the effect of each change.

The Study of Multiple Classes Boosting Classification Method Based on Local Similarity

Boosting of the ensemble learning model has made great progress, but most of the methods are Boosting the single mode. For this reason, based on the simple multiclass enhancement framework that uses local similarity as a weak learner, it is extended to multimodal multiclass enhancement Boosting. First, based on the local similarity as a weak learner, the loss function is used to find the basic loss, and the logarithmic data points are binarized. Then, we find the optimal local similarity and find the corresponding loss. Compared with the basic loss, the smaller one is the best so far. Second, the local similarity of the two points is calculated, and then the loss is calculated by the local similarity of the two points. Finally, the text and image are retrieved from each other, and the correct rate of text and image retrieval is obtained, respectively. The experimental results show that the multimodal multi-class enhancement framework with local similarity as the weak learner is evaluated on the standard data set and compared with other most advanced methods, showing the experience proficiency of this method.

A Novel Active Optimization Approach for Rapid and Efficient Design Space Exploration Using Ensemble Machine Learning

In this work, a novel design optimization technique based on active learning, which involves dynamic exploration and exploitation of the design space of interest using an ensemble of machine learning algorithms, is presented. In this approach, a hybrid methodology incorporating an explorative weak learner (regularized basis function model) that fits high-level information about the response surface and an exploitative strong learner (based on committee machine) that fits finer details around promising regions identified by the weak learner is employed. For each design iteration, an aristocratic approach is used to select a set of nominees, where points that meet a threshold merit value as predicted by the weak learner are selected for evaluation. In addition to these points, the global optimum as predicted by the strong learner is also evaluated to enable rapid convergence to the actual global optimum once the most promising region has been identified by the optimizer. This methodology is first tested by applying it to the optimization of a two-dimensional multi-modal surface and, subsequently, to a complex internal combustion (IC) engine combustion optimization case with nine control parameters related to fuel injection, initial thermodynamic conditions, and in-cylinder flow. It is found that the new approach significantly lowers the number of function evaluations that are needed to reach the optimum design configuration (by up to 80%) when compared to conventional optimization techniques, such as particle swarm and genetic algorithm-based optimization techniques.

Ensemble mobility predictor based on random forest and Markovian property using LBSN data

The ubiquitous connectivity of Location-Based Systems (LBS) allows people to share individual location-related data anytime. In this sense, Location-Based Social Networks (LBSN) provides valuable information to be available in large-scale and low-cost fashion via traditional data collection methods. Moreover, this data contains spatial, temporal, and social features of user activity, enabling a system to predict user mobility. In this sense, mobility prediction plays crucial roles in urban planning, traffic forecasting, advertising, and recommendations, and has thus attracted lots of attention in the past decade. In this article, we introduce the Ensemble Random Forest-Markov (ERFM) mobility prediction model, a two-layer ensemble learner approach, in which the base learners are also ensemble learning models. In the inner layer, ERFM considers the Markovian property (memoryless) to build trajectories of different lengths, and the Random Forest algorithm to predict the user’s next location for each trajectory set. In the outer layer, the outputs from the first layer are aggregated based on the classification performance of each weak learner. The experimental results on the real user trajectory dataset highlight a higher accuracy and f1-score of ERFM compared to five state-of-the-art predictors.

FUSION OF BAGGING BASED ENSEMBLE FRAMEWORK FOR EPILEPTIC SEIZURE CLASSIFICATION

The ensemble learning approach, especially in classification, has been widely carried out and is successful in many scopes, but unfortunately not many ensemble approaches are used for the detection and classification of epilepsy in biomedical terms. Compared to using a simple bagging ensemble framework, we propose a fusion bagging-based ensemble framework (FBEF) that uses 3 weak learners in each oracle, using fusion rules, a weak learner will give results as predictors of the oracle. All oracle predictors will be included in the trust factor to get a better prediction and classification. Compared to traditional Ensemble bagging and single learner type Ensemble bagging, our framework outperforms similar research in relation to the epileptic seizure classification as 98.11±0.68 and several real-world datasets

Pebandingan Metode K-Nearest Neighbor dan Adaptive Boosting pada Kasus Klasifikasi Multi Kelas

Kesehatan keuangan perusahaan memberikan suatu indikasi kinerja perusahaan yang berguna untuk mengetahui posisi perusahaan dalam area industri. Kinerja perusahaan perlu diprediksi untuk mengetahui perkembangan perusahaan. K-Nearest Neighbor (KNN) dan Adaptive Boosting (AdaBoost) merupakan metode klasifikasi yang dapat digunakan untuk memprediksi kinerja perusahaan. KNN mengklasifikasikan data berdasarkan kedekatan jarak data sedangkan AdaBoost bekerja dengan konsep memberi bobot lebih pada amatan yang termasuk weak learner. Tujuan dari penelitian ini adalah membandingkan metode KNN dan AdaBoost untuk mengetahui metode yang lebih baik dalam memprediksi kinerja perusahaan di Indonesia. Variabel dependen yang digunakan dalam penelitian ini adalah kinerja perusahaan yang digolongkan ke dalam empat kelas yaitu tidak sehat, kurang sehat, sehat, dan sehat sekali. Variabel independen yang digunakan terdiri atas tujuh rasio keuangan yaitu ROA, ROE, WCTA, TATO, DER, LDAR, dan ROI. Data yang digunakan yaitu data rasio keuangan dari 575 perusahaan yang tercatat di Bursa Efek Indonesia tahun 2019. Hasil penelitian ini menunjukkan bahwa prediksi kinerja perusahaan di Indonesia sebaiknya menggunakan metode AdaBoost karena memiliki akurasi klasifikasi sebesar 0,84522 yang lebih besar dibandingkan akurasi metode KNN sebesar 0,82087