Pattern Extraction Research Articles

Multi-stage degradation feature with dynamic feedback mechanism for remaining useful life prediction

ABSTRACT Bearings, essential in industrial rotating machinery, undergo progressive wear over time. This degradation occurs in multiple stages with varying rates, characteristic changes, and information content, presenting challenges for accurate remaining useful life (RUL) prediction. Existing methods do not account for the impact of multi-stage degradation characteristic differences on the prediction model, leading to significant prediction errors. This paper presents a novel dynamic feedback mechanism-driven approach based on multi-stage degradation characteristics. The method introduces a trend detection technique to segment monitoring data into distinct degradation stages. By continuously updating the parameters of the degradation model, the model accurately identifies transition points between stages. In a key innovation, a dynamic feedback mechanism is established between multi-stage features and multi-scale parameters, integrated with a multi-scale densely connected temporal convolutional network (TCN) prediction model. This mechanism enables adaptive adjustments to dilation rates and kernel sizes in response to feature variations across different stages, facilitating the precise extraction of stage-specific degradation patterns. Experimental validation using the XJTU-SY and SMU bearing datasets confirms that the proposed method outperforms existing approaches in terms of degradation feature extraction and multi-stage adaptive prediction, showcasing its practical applicability and superior performance in complex industrial settings..

Empirical Data-Driven Linear Model of a Swimming Robot Using the Complex Delay-Embedding DMD Technique.

Anguilliform locomotion, an efficient aquatic locomotion mode where the whole body is engaged in fluid-body interaction, contains sophisticated physics. We hypothesized that data-driven modeling techniques may extract models or patterns of the swimmers' dynamics without implicitly measuring the hydrodynamic variables. This work proposes empirical kinematic control and data-driven modeling of a soft swimming robot. The robot comprises six serially connected segments that can individually bend with the segmental pneumatic artificial muscles. Kinematic equations and relations are proposed to measure the desired actuation to mimic anguilliform locomotion kinematics. The robot was tested experimentally and the position and velocities of spatially digitized points were collected using QualiSys® Tracking Manager (QTM) 1.6.0.1. The collected data were analyzed offline, proposing a new complex variable delay-embedding dynamic mode decomposition (CDE DMD) algorithm that combines complex state filtering and time embedding to extract a linear approximate model. While the experimental results exhibited exotic curves in phase plane and time series, the analysis results showed that the proposed algorithm extracts linear and chaotic modes contributing to the data. It is concluded that the robot dynamics can be described by the linearized model interrupted by chaotic modes. The technique successfully extracts coherent modes from limited measurements and linearizes the system dynamics.

Prevalence of Reasons for Tooth Extraction in Small- and Medium-Breed Dogs.

As the life span of companion animals increases and their owners' concern for their quality of life increases, interest in dental diseases has also gained attention. While many studies have explored oral diseases in general, research on small-breed dogs remains limited. Therefore, in this study, we analyzed various diseases that led to tooth extraction in small- and medium-breed dogs. A study involving 2201 client-owned small- and medium-breed dogs was conducted. All dogs were categorized by sex, age, breed, and tooth position. Teeth were classified into untreated, missing, or extracted categories. For extracted teeth, 14 causes were identified. The results revealed that periodontal disease was the primary cause of tooth extractions, particularly in older dogs and small breeds. In addition to periodontal disease, various other diseases that led to tooth extraction were identified. For these other diseases, the number of extracted teeth increased with age, and brachycephalic breeds exhibited higher extraction rates. The study emphasizes the need for personalized oral care and further research into other contributing factors such as diet and genetics to better understand and manage oral health in dogs. It also offers valuable insights into the broader patterns of tooth extraction associated with various dental diseases.

Modulation pattern recognition method of wireless communication automatic system based on IABLN algorithm in intelligent system.

The aim of this study is to address the limitations of convolutional networks in recognizing modulation patterns. These networks are unable to utilize temporal information effectively for feature extraction and modulation pattern recognition, resulting in inefficient modulation pattern recognition. To address this issue, a signal modulation recognition method based on a two-way interactive temporal attention network algorithm has been developed. A two-way interactive temporal network is designed on the basis of the long and short-term memory network with the objective of enhancing the contextual connection of the temporal network. The output of the temporal network is attentively weighted using the soft attention mechanism. The proposed algorithm exhibited enhanced overall, average, and maximum recognition rates at varying signal-to-noise ratios, with an increase of 10.34%, 8.33%, and 3.33%, respectively, in comparison to other algorithms within the Radio Machine Learning (RML) 2016.10b dataset. Furthermore, the modulated signal recognition accuracy was as high as 92.84%, with an average increase in the Kappa coefficient of 12.28%. The Kappa coefficient in the Communication Signal Processing Benchmark for Machine Learning (CSPB.ML2018) 2018 dataset was 0.62, representing an average increase of 10.32% over other algorithms. The results demonstrate that the proposed recognition method can enhance the network's accuracy in recognizing modulated signals. Moreover, it has potential applications in modulation pattern recognition in automatic systems for wireless communications.

Herbicide and irrigation management options in conventionally-tilled wheat: deciphering water and energy budgeting, and grain and monetary output in north-Indian plains

In South Asia, declining water tables due to increased irrigation and labor shortages for manual weeding pose significant challenges for wheat production. Additionally, herbicide resistance, often resulting from poor management practices, further complicates weed problems. The objective of this study was to assess the impacts of traditional irrigation regimens (IRs) and herbicide application on wheat crops. The findings showed that when irrigation was applied at 100 mm CPE (IR4), and at 40 mm CPE (IR1), different combinations of herbicide to weed managment were tested. In comparison to the other treatments, application of irrigation at 40 mm cumulative pan evaporation (IR1) along with the Pendimethalin1000 g ha−1 (pre-em) in combination with clodinafop-propargyl 10% + metribuzin 22% + sulfosulfuron 4.2% at a rate of 1125 g ha−1 at 30 DAS (WM1) produced the best results in terms of crop yields, economic returns, relative water content, consumptive use, rate of water use, water use efficiency, water productivity, energy input–output, energy returns, energy productivity, energy intensity, specific energy, energy efficiency, maximum field capacity, available soil water, and soil profile moisture extraction pattern. The only exceptions were Pendimethalin1000g ha−1 (pre-em) combined with carfentrazone ethyl 20% + sulfosulfuron 25%WG), at the rate of 100 g ha−1 at 30 DAS (WM2) and the weed-free treatment (WM5), where the differences were not statistically significant. The yield of wheat grain (14.26 kg ha−1) and straw (14.41 kg ha−1) decreased as the unit dry matter production of weeds increased. The study recommends exploring additional weed control strategies and irrigation management options in future improve wheat yields in conventionally-tilled systems.

Credit Card Fraud Detection in the Banking Sector: A Comprehensive Machine Learning Approach for Information Security

In the context of computers, cybersecurity is experiencing significant technological development, and the changes have been driven by its operations in recent years. The secret to creating an intelligent and automated security system is to extract patterns or insights from cybersecurity data and create a matching data-driven model. One of the main problems, and threats to information security, is fraud. Credit card fraud detection (CCFD) is a significant issue for consumers, businesses, and banks, mostly because of the growth of computerized financial transactions. Because of this, a methodology for detecting fraud is presented that uses state-of-the-art machine learning (ML) techniques. The methodology in this research is a carefully chosen set of state-of-the-art ML algorithms that are particularly made for accurate CCFD problems. The technique uses a wide range of ML models to handle large-scale problems with a large number of transactions. Three ML models are used in this study, such as logistic regression (LR), random forest (RF), and XGBoost. These models are trained for accurate results of CCFD. Four evaluation metrics are used in this study to evaluate the ML models, such as accuracy, precision, recall, and f1 score. The results show that the RF model has the highest accuracy of 99.65%, followed by the XGBoost, with 99.963% accuracy, and the LR model, with 99.934% accuracy. The study's summary gives banking organizations, governmental organizations, and legislators crucial knowledge to help them fight against the harm that credit card theft does to customers, businesses, and the economy at large. By offering an ML-driven solution to the fraud problem, our work solves it and opens the door for further advancements in this important field.

Short-Term Load Forecasting Model Based on Time Series Clustering and Transformer in Smart Grid

Accurate Short-Term Load Forecasting (STLF) is a critical task in managing and operating smart grids. Existing STLF methods primarily rely on mathematical modeling or neural networks, often struggle to effectively capture the correlations between influencing factors and load data, and frequently lack interpretability. To address these challenges, this paper proposes an intelligent framework for STLF that combines a pattern extraction and attention mechanism, which leverages the characteristics of electricity consumption data. The proposed framework facilitates the integration of prior knowledge, identifies intrinsic data patterns, and more accurately maps the relationships between influencing factors and load patterns. Finally, we conduct experiments on real-world and publicly available datasets to evaluate the performance of the proposed model. Specifically, the proposed model improves the accuracy of STLF compared to that of existing methods and reduces the mean absolute percentage error by 2% to 5%. The model performs superiorly on the real datasets, with root mean squared error and mean absolute percentage error values of 0.810 MWh and 7.09%.

Male-assisted training and injury patterns: hypergraph-enhanced analysis of injuries in women's water polo.

The aim of this study is to compare the injury patterns of female water polo players before and after the implementation of the Male-Assisted Female Training (MAFT) program. The study seeks to identify key factors influencing these changes and propose corresponding injury prevention measures. We utilized pattern analysis and classification techniques to explore the injury data. A Hypergraph Neural Network (HGNN) was employed for pattern extraction, where each athlete was represented as a node in a hypergraph, with node dimensions capturing high-order relational embedding information. We applied the graph Laplacian operator to aggregate neighborhood features and visualize structural and feature differences in hypergraphs based on different influencing factors. Additionally, we introduced graph structure regularization to improve classification accuracy and prevent overfitting in the relatively small dataset, enhancing our ability to identify critical factors affecting injury types. The analysis revealed significant differences in injury patterns before and after the MAFT program, with specific influencing factors being identified through both pattern recognition and classification techniques. The classification models, supported by graph structure regularization, achieved improved accuracy in distinguishing key features that contributed to changes in injury types. These findings provide insights into the critical factors influencing injury patterns in female water polo players and highlight the effectiveness of the MAFT program in mitigating certain injury risks. Based on the identified features, we propose targeted preventive measures to reduce injury incidence, particularly in relation to changes brought about by the MAFT training mode. Further research is needed to refine these measures and explore their long-term effectiveness.

Hybrid CNN‐GNN Framework for Enhanced Optimization and Performance Analysis of Frequency‐Selective Surface Antennas

ABSTRACTFrequency‐selective surface (FSS) antennas are critical in modern communication systems, where optimizing their design for enhanced performance is essential. However, traditional methods often struggle with the complexity of FSS structures, leading to suboptimal designs. This paper addresses these limitations by proposing a novel CNN‐GNN hybrid network (CGHN) framework for FSS antenna optimization. The proposed methodology integrates convolutional neural networks (CNNs) for efficient feature extraction of spatial patterns within FSS designs and graph neural networks (GNNs) to model the relational dependencies between unit cells. This approach ensures that both local features and global interactions are captured, leading to more accurate and optimized antenna designs. The objective is to enhance the performance of FSS antennas by leveraging the complementary strengths of CNNs and GNNs, with an emphasis on improving design accuracy and efficiency. The novelty lies in the combination of CNN's localized pattern recognition with GNN's relational learning, which together enable a comprehensive understanding of the antenna's behavior. The proposed CGHN framework achieves a 96.78% accuracy rate in predicting optimal FSS designs, with a 23.84% boost in performance due to CNN‐driven feature extraction. Additionally, implementing stochastic gradient descent with gradient clipping increased the F1 score by 15%. Compared with existing techniques, the proposed method demonstrates significant improvements in both accuracy and efficiency, making it a superior choice for FSS antenna design optimization.

Modeling of the Extraction Kinetics of Oil from Safou Pulp (Dacryodes edulis H.J. Lam) by Heating in Hexane with Stirring

The continuous extraction of oil from safou pulp in hot hexane with stirring was studied. Two factors were considered: the method with two modes, M1 and M2, respectively without and with prior preheating of the solvent, and the temperature with two modes as well, 90 and 105°C for M1, and 50 and 70°C for M2. The results showed that continuous extraction by direct contact of the substrate with the solvent yields lower results compared to Soxhlet extraction, which has the advantage of regenerating the pure solvent each time, leading to more efficient recovery of the solute. However, the extraction rate is proportional to the preheating of the solvent, temperature, and agitation of the system, compared to previous results. The extraction curves exhibit the usual pattern of solid-liquid extraction of plants, with an ascending phase followed by stabilization. The temperature stabilization phase was not observed. The experimental curves validate the Peleg model, but not the first-order kinetic model. The oil is extracted by elution and diffusion. Peleg's lines, t/Yt = 0.01855t + 0.13074 at 90 °C and t/Yt = 0.02076t + 0.05839 at 105 °C for method M1, and t/Yt = 0.01842t + 0.03657 at 50 °C and t/Yt = 0.01934t + 0.02647 at 70 °C for method M2, characterizing the extraction process, confirm the previous conclusions. The initial extraction rates, respectively equal to 7.65, 17.13, 27.34, and 37.77 %.min-¹, are proportional to the extraction temperature and higher when the solvent is preheated. The activation energy values, 61.31 kJ.mol-¹ for method M1 and 14.88 kJ/mol for method M2, clearly show the benefit of preheating the solvent before starting the extraction. Int. J. Appl. Sci. Biotechnol. Vol 12(4): 224-234.

Layer porosity in powder-bed fusion prediction using regression machine learning models and time-series features

Additive manufacturing (AM) using laser powder-bed fusion (L-PBF) has become a common industrial process for high-end component production. The uptake of the process has been accelerated through the broad acceptance of the L-PBF process toward achieving high-quality parts with complex geometry. However, the L-PBF process faces challenges from the process’s sensitivity to the process build parameters, which, when incorrectly set, can cause defects such as porosity, which in turn have a detrimental effect on the produced part properties. On the other hand, the AM processing equipment generates a vast amount of data captured through in situ sensors such as pyrometers and imaging cameras. Having such an abundance of process data facilitates the employment of advanced machine learning (ML) tools to understand and extract patterns and information about the underlying AM process and gain “predictive control.” Driven by this idea, we aimed to employ ML tools over pyrometer time-series data from an L-PBF process to predict the porosity percentage of layers of an AM-built part. Sensor data are naturally modeled by time series; however, most ML algorithms work with tabular data (i.e., one single vector describes a feature). In the work presented here, feature engineering tools were used to transform the time-series data into informative features. These features were fed into the tabular ML algorithms for evaluation, broadening the selection of ML algorithms available in the literature. It was hypothesized that the time-series summary features would capture the interaction of melt-pool temperature with resulting porosity, from which the resulting models could better predict porosity occurrence. The dataset contains layer porosity values in the range of 0.00175 – 7.160%, to which we divide the data into “low” and “high” porous layers using a splitting threshold value of 1%. From evaluating these algorithms, it was concluded that classifying “low” versus “high” porosity layers is relatively easier than predicting the layer’s porosity percentage.

A study on identifying representative trips for mobility service design

AbstractRecently, with growing interest in urban mobility patterns, the demand for collecting and analysing origin‐destination (OD) data is increasing. Due to the large scale and dimensionality of OD data, there are two issues in analysing the data: big‐data storage and major pattern extraction. To deal with two issues at the same time, this study suggests a principal control analysis‐based major demand identification method to improve the usability of microscopic OD data. Especially, this study focuses on finding principal components that preserve major patterns from OD data with small random noise so that the data can be effectively used for mobility service design. The proposed method is applied to smart card data of Seoul and Sejong and extracted major demand patterns from peak‐ and non‐peak hour data of these cities. The degree of daily regularity, reconstruction accuracy, and compression rate of the reconstructed data is analysed varying sets of principal components. The obtained results show that the major demands contain a low volume and a large volume of demand and with lower‐order principal components, major demands can be efficiently extracted by removing randomly appearing small‐volume demand. In addition, the trade‐off behaviour is observed between the degree of daily regularity and reconstruction accuracy depending on the compression rate. Based on the observations, it can be found that the loss of major demand patterns could be prevented when targeting a reconstruction accuracy of 90–95% and the proposed method can reduce the data size while preserving major mobility patterns.

Differences in plant availability of phosphorus and potassium in pelletized and granulated biochar for an Andisol

AbstractBiochar will be broken into tiny pieces for processes of production, transportation, and application, which causes dust emissions for those processes. Particulate matter (PM) released from biochar may have negative effects on human health and increase the atmospheric burden of shortwave absorbing black carbon aerosols. Pelletizing feedstock before the thermochemical conversion is expected to reduce the emission of PM in the processing and post-processing phases. Effects of application of pelletized biochar, produced from broiler manure at different pyrolysis temperatures, into an Andisol on soil physicochemical properties and crop yields have been investigated in this study. Effects of pelletizing and pyrolysis temperatures (400, 600, and 800 °C) on dissolution properties and soil physicochemical properties were completely different between phosphorus and potassium. Pelletized broiler manure–derived biochar can be used as fast-release potassium fertilizer regardless of pyrolysis temperatures. In contrast, the extraction pattern of phosphorus contained in the biochar significantly differed by pelletizing and pyrolysis temperatures. Plant dry yields and phosphorus uptakes in soils amended with granulated broiler manure-derived biochar were significantly higher than those in soils amended with pelletized broiler manure-derived biochar (e.g., 3.4 times larger in 800 °C). This result suggests that the breakdown of pelletized biochar into granulated biochar could improve interaction between phosphorus contained in biochar and roots. Pelletized biochar will be transformed into granulated biochar through the freeze–thaw cycle, dry–wet cycle, and rotary tillage over the long term. Therefore, pelletized biochar derived from broiler manure at higher pyrolysis temperatures can be novel phosphorus-supplying amendments over the long term. Graphical abstract

TAXONOMIC DESCRIPTION AND ANTIBACTERIAL ACTIVITY OF DILLENIA sp. AGAINST ESCHERICHIA coli

Article Highlights- The study first reported taxonomic study of Dillenia sp. in Malita, Davao, Occidental Philippines. - This is the first reported findings of the bacterial activity of the bark extract of Dillenia sp. against Escherichia coli. AbstractThe study aimed to provide a taxonomic description of Dillenia sp. and evaluate its antibacterial activity against Escherichia coli (E. coli). The taxonomic description was characterized using the taxonomic keys and field guides. The antibacterial activity of Dillenia sp. was screened using the Kirby-Bauer disk diffusion assay. Taxonomic description confirms Dillenia sp. to be an angiosperm flowering seed plant native to the Philippines, characterized by its round-shaped growth and evergreen foliage. The antibacterial activity of Dillenia sp. bark extracts against E. coli were assessed using different concentrations (100 ppm, 500 ppm, and 1000 ppm) and extraction solvents (ethanolic, aqueous, and decoction). There are significant differences in the antibacterial activity observed among treatments within the decoction extracts, indicating varying effects on antibacterial activity. Post-hoc analysis revealed that concentrations of 100 ppm and 500 ppm were significantly more effective in inhibiting bacterial growth compared to 1000 ppm within the Decoction treatment. Each treatment showed distinct patterns of antibacterial activity, with ethanolic and aqueous extracts displaying relatively consistent activity across different concentrations, while the decoction extract exhibited concentration-dependent antibacterial activity. The study provides evidence of the antibacterial potential of Dillenia sp. bark extracts against Escherichia coli, with implications for further research and potential practical applications in combating bacterial infections. Recommendations include exploring different lower concentration ranges, considering different treatment formulations, investigating combination treatments, and assessing antimicrobial mechanisms to enhance efficacy and guide the development of novel antibacterial strategies.

Hybrid mechanism‐data‐driven iron loss modelling for permanent magnet synchronous motors considering multiphysics coupling effects

AbstractThe precise calculation of iron losses in permanent magnet synchronous motors (PMSMs) remains challenging due to the interplay between various disciplines such as electromagnetism, magnetism, and thermal/mechanical dynamics. Purely mechanistic models require detailed theoretical knowledge and exact parameters, often struggling to accurately describe complex systems, while purely data‐driven methods lack interpretability, which are susceptible to data noise and outliers in feature extraction and complicated pattern recognition. Consequently, this paper aims to present a hybrid mechanism‐data‐driven model for accurately estimating the iron loss for PMSMs, considering the multiphysics coupling effects. Specifically, based on the well‐defined physical principles, an advanced iron loss analytical model that simultaneously considers mechanical stress, temperature rise, harmonics, load currents, and changing frequency is developed and then utilised to calculate numerous loss data under different operating conditions, providing a certain level of stability and reliability for prediction accuracy. Subsequently, a convolutional neural network (CNN) algorithm is employed to perform deep learning to extract features and patterns from the data. By defining a suitable loss function, the pre‐trained model was fine‐tuned and optimised using a small amount of actual data. To validate its superiority, extensive numerical and experimental analyses are conducted on the prototype. The results demonstrate that the iron losses computed using this hybrid model overcome the limitations of singular methods by effectively leveraging both theoretical knowledge and real‐world data, thus accurately accommodating various application scenarios. This integrated approach enhances the accuracy, stability, and interpretability of the model, laying a solid foundation for more specialised applications in the future.

Quantifying the Contribution and Mobility of In Situ-Produced Helium in He-Retentive Minerals: A Case Study of the Salla-Kuolajarvi Metasomatic Rocks (Russia)

An alternative approach to separating trapped and radiogenic helium, and assessing the mobility of the latter in He-retentive minerals implemented in the present work, has been developed on the basis of helium extraction patterns obtained via incremental heating of mineral samples. We used these data both to estimate helium mobility in concentrates of various ore minerals (pyrite, magnetite, and hematite) and to assess the contribution of in situ-produced 4He in the total helium budget. The rocks of uranium-ore and gold mineralization of the Salla-Kuolajarvi belt (northern Karelia, Fennoscandian Shield) were used as test materials for the new approach. The method allowed such parameters as diffusivity, activation energy, closure temperature, and the contribution of the trapped helium to be obtained for all samples. The latter was used for the correction of apparent U-Th-He bulk ages. The interpretation of the calculated values was performed taking into account the closure temperatures of the U-Th-He system, as well as the peculiarities of each individual sample.

HCLNet: A Hybrid CNN-LSTM Model for Enhanced Healthcare Product Classification for Recommendation

The rapid growth of healthcare commodities requires automated categorization. Effective categorization solutions save time and money and can provide precise recommendations. Poorly annotated data, inconsistency of product category, and imbalanced datasets impede machine learning algorithms in current systems. CNNs work well with large datasets but struggle with the classification of healthcare products due to class imbalance and lack of labeled data. This study presents a Hybrid CNN-LSTM (HCLNet) model to improve the classification of healthcare products and resolve these issues. This method enhances classification using CNN feature extraction and LSTM sequential pattern recognition. HCLNet can better handle class imbalance and limited labeled data with a comprehensive data preprocessing pipeline, including selection, transformation, and filtering. The hybrid design overcomes CNN constraints and captures product feature temporal connections using LSTM layers. HCLNet was compared with ResNet, GoogleNet, and AlexNet, surpassing them. HCLNet classified complex and imbalanced datasets with 96.25% accuracy, 96.60% precision, and 96.05% recall. The proposed method can improve the classification of healthcare products to obtain accurate automated product recommendations.

Chemometrics and digital image colorimetry approaches applied to paper-based analytical devices: A review.

Colorimetric paper-based analytical devices (CPADs) are cost-efficient and high-throughput technologies that use readily available materials for point-of-need (PON) applications by leveraging color changes in response to target analytes. However, the complexity of samples can limit the precision and accuracy of CPAD applications. Therefore, CPADs have been combined with chemometric approaches to enhance analytical performance and provide simple solutions to complex systems. The integration of formal optimization techniques, such as Design of Experiments (DoE), classification tools, quantification methods, and other advanced algorithms enables optimal experimental analytical conditions and extracting meaningful information from the complex colorimetric data generated by CPADs. These approaches facilitate robust calibration and prediction models, enabling reliable quantifications or sample differentiation. In addition, chemometrics combined with CPADs contributes to Green Analytical Chemistry once they have the potential to minimize the number of experiments, provide optimal designs and consumption of reagents, and decrease waste generation. The synergy between digital colorimetry using CPADs and smart devices with chemometric techniques holds great promise for portable analysis in resource-limited settings, with applications ranging from environmental monitoring to point-of-care diagnostics. Herein, we review recent advances in the development of CPADs, ranging from manufacturing methods to extraction of color patterns and data treatment using several chemometric tools, performance assessment, and potential transfer to onsite applications for relevant analytical problems.

Efficient Integration of CNN and LSTM for Data Mining Observation Algorithms: A Systematic Review

An extensive analysis of the combination of Long Short-Term Memory (LSTM) networks with Convolutional Neural Networks (CNN) is presented in this research. within data mining applications. CNNs, optimized for spatial data processing, and LSTMs, made to manage data in a sequential fashion, are combined to leverage their respective strengths in feature extraction and temporal pattern recognition. By analyzing over 60 published studies, we evaluate the effectiveness of CNN-LSTM models across diverse domains such as medical image analysis, financial time series forecasting, stock price prediction, and sentiment analysis. The review reveals that CNN-LSTM hybrids significantly outperform traditional methods in activities requiring both temporally and spatially understanding. However, challenges such as model interpretability, hyperparameter optimization, and the need for diverse datasets remain. Future research should focus on optimizing these models for specific applications, improving scalability, and addressing limitations like data scarcity. This review underscores the significant potential of CNNLSTM integration in advancing data mining techniques and solving complex problems in various fields. We present the related work of some of the most important publication in the field and classify over 60 published papers from different perspective. This review intends to classify the major techniques, methods, algorithms in the domain of medical images, financial time series analysis, stock price and social media.

Extraction of patterns from images using a model of combined frequency localization spaces

An algorithm for image decomposition and separation of superposed stationary contributions is proposed. It is based on the concept of sparse-to-sparse domain representation achieved through a relationship between block-based and full-size discrete cosine transform. The L-statistics is adapted to discard nonstationary components from the frequency domain vectors, leaving just a few coefficients associated with stationary pattern. These fewer stationary components are then used under the compressive sensing framework to reconstruct the stationary pattern. The original image is observed as a nonstationary component, acting as a non-desired part at this stage of the procedure, while the stationary pattern is observed as a “desired part” that should be extracted through the reconstruction process. The problem of interest is formulated as underdetermined system of equations resulting from a relationship between the two considered transformation spaces. Once the stationary pattern is reconstructed, it can be removed entirely from the image. Furthermore, it will be shown that the efficiency of pattern extraction cannot be affected, even when image contains additional nonstationary disturbance (here, the noisy image is observed as nonstationary undesired part). The proposed approach is motivated by challenges in removing Moiré-like patterns from images, enabling some interesting applications, including extraction of hidden sinusoidal signatures.