• Journal of Navigation and Port Research
• /
• v.47 no.6
• /
• pp.376-385
• /
• 2023

Fishing boat capsizing accidents account for more than half of all capsize accidents. These can occur for a variety of reasons, including inexperienced operation, bad weather, and poor maintenance. Due to the size and influence of the industry, technological complexity, and regional diversity, fishing ships are relatively under-researched compared to commercial ships. This study aimed to predict the rolling motion time series of fishing boats using an image-based deep learning model. Image-based deep learning can achieve high performance by learning various patterns in a time series. Three image-based deep learning models were used for this purpose: Xception, ResNet50, and CRNN. Xception and ResNet50 are composed of 177 and 184 layers, respectively, while CRNN is composed of 22 relatively thin layers. The experimental results showed that the Xception deep learning model recorded the lowest Symmetric mean absolute percentage error(sMAPE) of 0.04291 and Root Mean Squared Error(RMSE) of 0.0198. ResNet50 and CRNN recorded an RMSE of 0.0217 and 0.022, respectively. This confirms that the models with relatively deeper layers had higher accuracy.

• Tunnel and Underground Space
• /
• v.28 no.6
• /
• pp.651-669
• /
• 2018

This study suggests the prediction model to estimate the specific energy of a pick cutter using a gene expression programming (GEP) and particle swarm optimization (PSO). Estimating the performance of mechanical excavators is of crucial importance in early design stage of tunnelling projects, and the specific energy (SE) based approach serves as a standard performance prediction procedure that is applicable to all excavation machines. The purpose of this research, is to investigate the relationship between UCS and BTS, penetration depth, cut spacing, and SE. A total of 46 full-scale linear cutting test results using pick cutters and different values of depth of cut and cut spacing on various rock types was collected from the previous study for the analysis. The Mean Squared Error (MSE) associated with the conventional Multiple Linear Regression (MLR) method is more than two times larger than the MSE generated by GEP-PSO algorithm. The $R^2$ value associated with the GEP-PSO algorithm, is about 0.13 higher than the $R^2$ associated with MLR.

• Journal of the Korean Society of Marine Environment & Safety
• /
• v.28 no.4
• /
• pp.469-476
• /
• 2022

This study developed a Recurrent Neural Network model implemented through Long Short-Term Memory (LSTM) that generates long-term tidal level data at Busan Port using tide observation data. The tide levels in Busan Port were predicted by the Korea Hydrographic and Oceanographic Administration (KHOA) using the tide data observed at Busan New Port and Tongyeong as model input data. The model was trained for one month in January 2019, and subsequently, the accuracy was calculated for one year from February 2019 to January 2020. The constructed model showed the highest performance with a correlation coefficient of 0.997 and a root mean squared error of 2.69 cm when the tide time series of Busan New Port and Tongyeong were inputted together. The study's finding reveal that long-term tidal level data prediction of an arbitrary port is possible using the deep learning recurrent neural network model.

• International Journal of Computer Science & Network Security
• /
• v.23 no.9
• /
• pp.1-7
• /
• 2023

Demand prediction is an essential component of any business or supply chain. Large retailers need to keep track of tens of millions of items flows each day to ensure smooth operations and strong margins. The demand prediction is in the epicenter of this planning tornado. For business processes in retail companies that deal with a variety of products with short shelf life and foodstuffs, forecast accuracy is of the utmost importance due to the shifting demand pattern, which is impacted by an environment of dynamic and fast response. All sectors strive to produce the ideal quantity of goods at the ideal time, but for retailers, this issue is especially crucial as they also need to effectively manage perishable inventories. In light of this, this research aims to show how Machine Learning approaches can help with demand forecasting in retail and future sales predictions. This will be done in two steps. One by using historic data and another by using open data of weather conditions, fuel, Consumer Price Index (CPI), holidays, any specific events in that area etc. Several machine learning algorithms were applied and compared using the r-squared and mean absolute percentage error (MAPE) assessment metrics. The suggested method improves the effectiveness and quality of feature selection while using a small number of well-chosen features to increase demand prediction accuracy. The model is tested with a one-year weekly dataset after being trained with a two-year weekly dataset. The results show that the suggested expanded feature selection approach provides a very good MAPE range, a very respectable and encouraging value for anticipating retail demand in retail systems.

• International Journal of Computer Science & Network Security
• /
• v.23 no.8
• /
• pp.210-216
• /
• 2023

Demand prediction is an essential component of any business or supply chain. Large retailers need to keep track of tens of millions of items flows each day to ensure smooth operations and strong margins. The demand prediction is in the epicenter of this planning tornado. For business processes in retail companies that deal with a variety of products with short shelf life and foodstuffs, forecast accuracy is of the utmost importance due to the shifting demand pattern, which is impacted by an environment of dynamic and fast response. All sectors strive to produce the ideal quantity of goods at the ideal time, but for retailers, this issue is especially crucial as they also need to effectively manage perishable inventories. In light of this, this research aims to show how Machine Learning approaches can help with demand forecasting in retail and future sales predictions. This will be done in two steps. One by using historic data and another by using open data of weather conditions, fuel, Consumer Price Index (CPI), holidays, any specific events in that area etc. Several machine learning algorithms were applied and compared using the r-squared and mean absolute percentage error (MAPE) assessment metrics. The suggested method improves the effectiveness and quality of feature selection while using a small number of well-chosen features to increase demand prediction accuracy. The model is tested with a one-year weekly dataset after being trained with a two-year weekly dataset. The results show that the suggested expanded feature selection approach provides a very good MAPE range, a very respectable and encouraging value for anticipating retail demand in retail systems.

• Computers and Concrete
• /
• v.31 no.2
• /
• pp.85-95
• /
• 2023

Natural pozzolans are used as additives in cement to develop more durable and high-performance concrete. Pozzolanic activity index (PAI) is important for assessing the performance of a pozzolan as a binding material and has an important effect on the compressive strength, permeability, and chemical durability of concrete mixtures. However, the determining of the 28 days (short term) and 90 days (long term) PAI of concrete mixtures is a time-consuming process. In this study, to reduce extensive experimental work, it is aimed to predict the short term and long term PAIs as a function of the chemical compositions of various natural pozzolans. For this purpose, the chemical compositions of various natural pozzolans from Central Anatolia were determined with X-ray fluorescence spectroscopy. The mortar samples were prepared with the natural pozzolans and then, the short term and the long term PAIs were calculated based on compressive strength method. The effect of the natural pozzolans' chemical compositions on the short term and the long term PAIs were evaluated and the PAIs were predicted by using multiple linear regression (MLR) and adaptive neuro-fuzzy inference system (ANFIS) model. The prediction model results show that both reactive SiO₂ and SiO₂+Al₂O₃+Fe₂O₃ contents are the most effective parameters on PAI. According to the performance of prediction models determined with metrics such as root mean squared error (RMSE) and coefficient of correlation (R²), ANFIS models are more feasible than the multiple regression model in predicting the 28 days and 90 days pozzolanic activity. Estimation of PAIs based on the chemical component of natural pozzolana with high-performance prediction models is going to make an important contribution to material engineering applications in terms of selection of favorable natural pozzolana and saving time from tedious test processes.

• 전자공학회논문지 IE
• /
• v.48 no.2
• /
• pp.19-25
• /
• 2011

This paper presented neural network modeling method using vehicle data to predict fuel consumption. To acquire data for training and testing the proposed neural network, medium-class gasoline vehicle drove at downtown and parameters measured include speed, engine rpm, throttle position sensor (TPS), and mass air flow (MAF) as input data, and fuel consumption as target data from OBD-II port. Multi layer perception network was used for nonlinear mapping between the input and the output data. It was observed that the neural network model can predict the vehicle quite well with mean squared error was $1.306{\times}10^{-6}$ for the fuel consumption.

• Industrial Engineering and Management Systems
• /
• v.9 no.3
• /
• pp.262-274
• /
• 2010

This article investigates how to adaptively predict the time-varying metrology delay that could realistically occur in the semiconductor manufacturing practice. Metrology delays pose a great challenge for the existing run-to-run (R2R) controllers, driving the process output significantly away from target if not adequately predicted. First, the expected asymptotic double exponentially weighted moving average (DEWMA) control output, by using the EWMA and recursive least squares (RLS) prediction methods, is derived. It has been found that the relationships between the expected control output and target in both estimation methods are parallel, and six cases are addressed. Within the context of time-varying metrology delay, this paper presents a modified recursive least squares-linear trend (RLS-LT) controller, in combination with runs test. Simulated single input-single output (SISO) R2R processes subject to various time-varying metrology delay scenarios are used as a testbed to evaluate the proposed algorithms. The simulation results indicate that the modified RLS-LT controller can yield the process output more accurately on target with smaller mean squared error (MSE) than the original RLSLT controller that only deals with constant metrology delays.

• Advances in Energy Research
• /
• v.6 no.2
• /
• pp.91-101
• /
• 2019

This paper proposes an efficient data-driven approach to build models for predicting energy consumption in buildings. Data used in this research is collected by installing humidity and temperature sensors at different locations in a building. In addition to this, weather data from nearby weather station is also included in the dataset to study the impact of weather conditions on energy consumption. One of the main emphasize of this research is to make feature selection independent of domain knowledge. Therefore, to extract useful features from data, two different approaches are tested: one is feature selection through principal component analysis and second is relative importance-based feature selection in original domain. The regression model used in this research is gradient boosting regression and its optimal parameters are chosen through a two staged coarse-fine search approach. In order to evaluate the performance of model, different performance evaluation metrics like r2-score and root mean squared error are used. Results have shown that best performance is achieved, when relative importance-based feature selection is used with gradient boosting regressor. Results of proposed technique has also outperformed the results of support vector machines and neural network-based approaches tested on the same dataset.

• Journal of the military operations research society of Korea
• /
• v.33 no.1
• /
• pp.59-74
• /
• 2007

Computer simulation models are in general quite complex and time-consuming to run, and therefore, a simpler meta-model is usually constructed for further analysis. In this paper, JANUS, a war-game simulator, is used to describe a certain tank combat situation. Then, second-order response surface and artificial neural network meta-models are developed using the data from eight different experimental designs. Relative performances of the developed meta-models are compared in terms of the mean squared error of prediction. Computational results indicate that, for the given problem, the second-order response surface meta-model generally performs better than the neural network, and the orthogonal array-based Latin hypercube design(LHD) or LHD using maximin distance criterion may be recommended.