• International Journal of High-Rise Buildings
• /
• v.9 no.4
• /
• pp.335-342
• /
• 2020

This article comes up with a new method which is based on the visual characteristic of the objects and machine learning technology to achieve semi-automated recognition of the personnel, machine & materials of the construction sites. Balancing the real-time performance and accuracy, using Faster RCNN (Faster Region-based Convolutional Neural Networks) with transfer learning method appears to be a rational choice. After fine-tuning an ImageNet pre-trained Faster RCNN and testing with it, the result shows that the precision ratio (mAP) has so far reached 67.62%, while the recall ratio (AR) has reached 56.23%. In other word, this recognizing method has achieved rational performance. Further inference with the video of the construction of Huoshenshan Hospital also indicates preliminary success.

• The Journal of Bigdata
• /
• v.5 no.2
• /
• pp.215-229
• /
• 2020

The interest in machine learning is growing in all industries, but it is difficult to apply it to real-world tasks because of inexplicability. This paper introduces a case of developing a financial customer churn prediction model for a securities company, and introduces the research results on an attempt to develop a machine learning model that can be explained using the SHAP Value methodology and derivation of interpretability. In this study, a total of six customer churn models are compared and analyzed, and the cause of customer churn is inferred through the classification and data analysis of SHAP Value and the type of customer asset change. Based on the results of this study, it would be possible to use it as a basis for comprehensive judgment, such as using the Value of the deviation prediction result that can infer the cause of the marketing manager's actual customer marketing in the future and establishing a target marketing strategy for each customer.

• Geophysics and Geophysical Exploration
• /
• v.26 no.2
• /
• pp.62-76
• /
• 2023

Recently, machine learning (ML) techniques have been actively applied for seismic trace interpolation. However, because most research is based on training-inference strategies that treat missing trace gather data as a 2D image with a blank area, a sufficient number of fully sampled data are required for training. This study proposes trace interpolation using ML, which uses only irregularly sampled field data, both in training and inference, by modifying the training-inference strategies of trace-based interpolation techniques. In this study, we describe a method for constructing networks that vary depending on the maximum number of consecutive gaps in seismic field data and the training method. To verify the applicability of the proposed method to field data, we applied our method to time-migrated seismic data acquired from the Vincent oilfield in the Exmouth Sub-basin area of Western Australia and compared the results with those of the conventional trace interpolation method. Both methods showed high interpolation performance, as confirmed by quantitative indicators, and the interpolation performance was uniformly good at all frequencies.

• Geomechanics and Engineering
• /
• v.37 no.4
• /
• pp.307-321
• /
• 2024

This paper delves into the critical assessment of predicting sidewall displacement in underground caverns through the application of nine distinct machine learning techniques. The accurate prediction of sidewall displacement is essential for ensuring the structural safety and stability of underground caverns, which are prone to various geological challenges. The dataset utilized in this study comprises a total of 310 data points, each containing 13 relevant parameters extracted from 10 underground cavern projects located in Iran and other regions. To facilitate a comprehensive evaluation, the dataset is evenly divided into training and testing subset. The study employs a diverse array of machine learning models, including recurrent neural network, back-propagation neural network, K-nearest neighbors, normalized and ordinary radial basis function, support vector machine, weight estimation, feed-forward stepwise regression, and fuzzy inference system. These models are leveraged to develop predictive models that can accurately forecast sidewall displacement in underground caverns. The training phase involves utilizing 80% of the dataset (248 data points) to train the models, while the remaining 20% (62 data points) are used for testing and validation purposes. The findings of the study highlight the back-propagation neural network (BPNN) model as the most effective in providing accurate predictions. The BPNN model demonstrates a remarkably high correlation coefficient (R2 = 0.99) and a low error rate (RMSE = 4.27E-05), indicating its superior performance in predicting sidewall displacement in underground caverns. This research contributes valuable insights into the application of machine learning techniques for enhancing the safety and stability of underground structures.

• International Journal of Fuzzy Logic and Intelligent Systems
• /
• v.10 no.4
• /
• pp.314-318
• /
• 2010

Predicting Alpha-helicies, Beta-sheets and Turns of a proteins secondary structure is a complex non-linear task that has been approached by several techniques such as Neural Networks, Genetic Algorithms, Decision Trees and other statistical or heuristic methods. This project introduces a new machine learning method by combining Bayesian Inference with offline trained Multilayered Perceptron (MLP) models as the likelihood for secondary structure prediction of proteins. With varying window sizes of neighboring amino acid information, the information is extracted and passed back and forth between the Neural Net and the Bayesian Inference process until the posterior probability of the secondary structure converges.

• Geomechanics and Engineering
• /
• v.34 no.6
• /
• pp.697-726
• /
• 2023

Brittleness as an important property of rock plays a crucial role both in the failure process of intact rock and rock mass response to excavation in engineering geological and geotechnical projects. Generally, rock brittleness indices are calculated from the mechanical properties of rocks such as uniaxial compressive strength, tensile strength and modulus of elasticity. These properties are generally determined from complicated, expensive and time-consuming tests in laboratory. For this reason, in the present research, an attempt has been made to predict the rock brittleness indices from simple, inexpensive, and quick laboratory test results namely dry unit weight, porosity, slake-durability index, P-wave velocity, Schmidt rebound hardness, and point load strength index using multiple linear regression, exponential regression, support vector machine (SVM) with various kernels, generating fuzzy inference system, and regression tree ensemble (RTE) with boosting framework. So, this could be considered as an innovation for the present research. For this purpose, the number of 39 rock samples including five igneous, twenty-six sedimentary, and eight metamorphic were collected from different regions of Iran. Mineralogical, physical and mechanical properties as well as five well known rock brittleness indices (i.e., B₁, B₂, B₃, B₄, and B₅) were measured for the selected rock samples before application of the above-mentioned machine learning techniques. The performance of the developed models was evaluated based on several statistical metrics such as mean square error, relative absolute error, root relative absolute error, determination coefficients, variance account for, mean absolute percentage error and standard deviation of the error. The comparison of the obtained results revealed that among the studied methods, SVM is the most suitable one for predicting B₁, B₂ and B₅, while RTE predicts B₃ and B₄ better than other methods.

• Journal of the Regional Association of Architectural Institute of Korea
• /
• v.20 no.6
• /
• pp.113-119
• /
• 2018

Occupancy-based heating control is effective in reducing heating energy by preventing unnecessary heating during unoccupied period. Various technologies on detecting human occupancy have been developed using complicated machine learning algorithm and stochastic methodologies. This study aims at deriving low-cost and simple algorithm of occupancy inference that can be implemented to residential buildings. The core concept of the algorithm is to combine the occupancy probabilities based on indoor CO2 concentration and PIR(passive infrared) signals. The probability was estimated by applying different levels of decrement ratio depending on CO2 concentration change rate and aggregated PIR signals. The developed algorithm was validated by comparing the inference results with the occupancy schedule in a real residential building. The results showed that the inference algorithm can achieve the accuracy of 75~99%, which would be successfully implemented to the control of residential heating systems.

• Journal of the Korean BIBLIA Society for library and Information Science
• /
• v.28 no.4
• /
• pp.301-319
• /
• 2017

The task of author name disambiguation involves identifying an author with different names or different authors with the same name. The author name disambiguation is important for correctly assessing authors' research achievements and finding experts in given areas as well as for the effective operation of scholarly information services such as citation indexes. In the study, we performed error correction and normalization of data and applied rules-based author name disambiguation to compare with baseline machine learning disambiguation in order to see if human intervention could improve the machine learning performance. The improvement of over 0.1 in F-measure by the corrected and normalized email-based author name disambiguation over machine learning demonstrates the potential of human pattern identification and inference, which enabled data correction and normalization process as well as the formation of the rule-based diambiguation, to complement the machine learning's weaknesses to improve the author name disambiguation results.

• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 2003.09a
• /
• pp.480-483
• /
• 2003

We propose a new framework that aims at multi-purpose image recognition, a difficult task for the conventional rule-based systems. This framework is farmed based on the idea of computer-based learning algorithm. In this research, we introduce the new functions of an additional learning and a knowledge reconstruction on the Fuzzy Inference Neural Network (FINN) (1) to enable the system to accommodate new objects and enhance the accuracy as necessary. We examine the capability of the proposed framework using two examples. The first one is the capital letter recognition task from UCI machine learning repository to estimate the effectiveness of the framework itself, Even though the whole training data was not given in advance, the proposed framework operated with a small loss of accuracy by introducing functions of the additional learning and the knowledge reconstruction. The other is the scenery image recognition. We confirmed that the proposed framework could recognize images with high accuracy and accommodate new object recursively.

• IEMEK Journal of Embedded Systems and Applications
• /
• v.19 no.3
• /
• pp.151-158
• /
• 2024

Although on-device artificial intelligence (AI) has gained attention to diagnosing machine faults in real time, most previous studies did not consider the model retraining and redeployment processes that must be performed in real-world industrial environments. Our study addresses this challenge by proposing an on-device AI-based real-time machine fault diagnosis system that utilizes continual learning. Our proposed system includes a lightweight convolutional neural network (CNN) model, a continual learning algorithm, and a real-time monitoring service. First, we developed a lightweight 1D CNN model to reduce the cost of model deployment and enable real-time inference on the target edge device with limited computing resources. We then compared the performance of five continual learning algorithms with three public bearing fault datasets and selected the most effective algorithm for our system. Finally, we implemented a real-time monitoring service using an open-source data visualization framework. In the performance comparison results between continual learning algorithms, we found that the replay-based algorithms outperformed the regularization-based algorithms, and the experience replay (ER) algorithm had the best diagnostic accuracy. We further tuned the number and length of data samples used for a memory buffer of the ER algorithm to maximize its performance. We confirmed that the performance of the ER algorithm becomes higher when a longer data length is used. Consequently, the proposed system showed an accuracy of 98.7%, while only 16.5% of the previous data was stored in memory buffer. Our lightweight CNN model was also able to diagnose a fault type of one data sample within 3.76 ms on the Raspberry Pi 4B device.