Narrow Results

In the era of digital transformation, leveraging multi-source sensor data to analyze and enhance student learning behavior has become increasingly crucial. In this research, we propose a Learn-Sync Intelligent Fusion (LSIF) system that delves into Estimate Adam driven- Intelligent Gradient Boosting Machines (EA-IGBM) that integrate sensor data from diverse sources to provide extensive analysis for the student behavior based on engagement and performance using multi-source sensor data fusion. This research employs the sensor data for online and offline classes for predicting the student learning behaviors during the class. To provide a holistic view of student engagement and performance, these sensor data are normalized using min–max normalization. Recursive Feature Elimination (RFE) is employed to extract the normalized multi-source data to frame multiple features that are integrated using the feature-level fusion technique. The LSIF aggregates behavioral fusion data from online and offline activities, using the EA-IGBM algorithm to predict academic success and provide actionable feedback. This model stimulates using tensor flow 2.15. The proposed EA-IGBM algorithm for both offline and online learning engagement detection performance is more significant, which utilizes certain parameters like student engagement, interaction frequency, behavioral patterns, and distractive levels. The established system highlights the effectiveness of multi-source sensor data fusion in monitoring and optimizing student learning behavior.

Crack identification of buildings using the Internet of things (IoT) is done by continuously monitoring the building structures that provide an early indication of cracks in buildings. The established IoT system constantly gathers structural information using sensors and stores it on a cloud server. This paper presented an innovative machine learning crack identification methodology for detecting cracks using the sensor data. Initially, the collected sensor data is pre-processed by the cloud server using the data fusion process for further processing. Subsequently, effective damage sensitive features such as mode structure (MS) features such as damage signature, streamlined damage signature index, modal assurance criterion (MAC) and coordinate MAC, improved natural frequency (INF) features, and mode structure curvature (MSC) features with curvature damage factor are extracted from the pre-processed data to differentiate cracks easily. After features are extracted, the feature score-based random projection (FSRP) technique is utilized for dimensionality reduction. Finally, hybridization of improved convolutional neural network with modified whale optimization (ICNN-MWO) detects the cracks in the civil structure utilizing the selected features. These effective classification results might alert the user when a high severity or damage is likely to occur. The implementation platform used in this work is PYTHON. The experimental outcomes of the presented technique proved that the presented work is significantly better in terms of various effective performance measures like accuracy (99.93%), mean squared error (3%), precision (99.91%), recall (99.90%), and F-measure (99.9%). The experimental results of the presented methodology provide improved performance than the existing crack identification techniques.

New policies are commenced all over the globe to diminish the use of fossil fuels, which gives rise to the augmented utilization of solar energy (SE). The photovoltaic (PV) system’s performance is extremely environmental variables reliant. Long-range transmission of SE is incompetent as well as complex to carry in the PV system. It can be affected by disparate sorts of faults, which cause severe energy loss all through the system operation. Thus, it is vital to incessantly monitor the solar PV (SPV) system to detect as well classify the faults by preventing energy losses. The IoT applications in SE production engage sensor devices that are fixed to the generation, and transmission, together with distribution equipment. These devices assist in monitoring the operation of the SPV power plant (SPVPP) system remotely in real-time. Presenting a new algorithm that can perform fault detection and classification in a PV system to a higher level of accuracy is the major contribution of this work. Thus, this work designs as well as develops an IoT platform for carrying out analytical tasks that can analyze data generated as of IoT operating systems to detect as well as classify faults in the SPVPP. Initially, the data collected from the dataset is pre-processed in which data duplication is performed using Hadoop distributed file system (HDFS) and then the fault is detected from the pre-processed data using the cosine function based k-means clustering (CFKC) technique in the SPV system. Finally, the obtained fault data is fed into the optimized deep learning centered ENN (ODENN) method which classifies the faults. The proposed techniques detect as well as classify the faults effectively that are experimentally proved by means of comparing them with the prevailing techniques, namely ENN, ANN and SVM, along with KNN in terms of some quality measures. The obtained results for ODENN showed an accuracy of 98.99%, specificity of 97.6%, and a sensitivity of 97.02%.

In this paper, a fault prediction method for oil well equipment based on the analysis of time series data obtained from multiple sensors is proposed. The proposed method is based on deep learning (DL). For this purpose, comparative analysis of single-layer long short-term memory (LSTM) with the convolutional neural network (CNN) and stacked LSTM methods is provided. To demonstrate the efficacy of the proposed method, some experiments are conducted on the real data set obtained from eight sensors installed in oil wells. In this paper, compared to the single-layer LSTM model, the CNN and stacked LSTM predicted the faulty time series with a minimal loss.

This paper reports on how to transform a multiclass classification problem into a set of simpler classification problems and then combine the solutions to the simpler problems into a solution to the original multiclass classification. A novel two-layer framework is presented, called Hierarchical Classification Learning. Different machine learning algorithms can be employed as the base classifier in this classification learning framework. First of all, the multiclass data set is reformed for every pair of classes, resulting in multiple 3-class sensor data sets — two classes for the pair of classes and the third class for any data instance that not belong to any of the two classes. A classification model, called sensor model, is constructed for each of the sensor data sets using a machine learning algorithm, or the base classifier. Then every data instance of the original data set is put into all the sensor models, generating a set of sensor outputs or secondary features which compose a sensor vector. Every sensor vector is viewed as a reformed version of the original data that has the same class label as the original, so put all sensor vectors together and get a new reformed data set. A classification model, decision model, is constructed from the new reformed data set. At last, extensive experiments have been conducted to evaluate the framework, and neural network, decision tree, random tree, and support vector machine are used as the base classifiers. Experiment results on UCI datasets and some popular face classification datasets show that the classification learning framework has achieved superior performance than their corresponding base classifiers.

We propose hypothesis tests for detecting dopaminergic medication response in Parkinson disease patients, using longitudinal sensor data collected by smartphones. The processed data is composed of multiple features extracted from active tapping tasks performed by the participant on a daily basis, before and after medication, over several months. Each extracted feature corresponds to a time series of measurements annotated according to whether the measurement was taken before or after the patient has taken his/her medication. Even though the data is longitudinal in nature, we show that simple hypothesis tests for detecting medication response, which ignore the serial correlation structure of the data, are still statistically valid, showing type I error rates at the nominal level. We propose two distinct personalized testing approaches. In the first, we combine multiple feature-specific tests into a single union-intersection test. In the second, we construct personalized classifiers of the before/after medication labels using all the extracted features of a given participant, and test the null hypothesis that the area under the receiver operating characteristic curve of the classifier is equal to 1/2. We compare the statistical power of the personalized classifier tests and personalized union-intersection tests in a simulation study, and illustrate the performance of the proposed tests using data from mPower Parkinsons disease study, recently launched as part of Apples ResearchKit mobile platform. Our results suggest that the personalized tests, which ignore the longitudinal aspect of the data, can perform well in real data analyses, suggesting they might be used as a sound baseline approach, to which more sophisticated methods can be compared to.