جستجوی مقالات مرتبط با کلیدواژه "regression model" در نشریات گروه "فنی و مهندسی"

In the stock market, which is a dynamic, complex, nonlinear and non-parametric environment, accurate prediction is crucial for trading strategy. It is assumed that news articles affect the stock market. We investigated the relationship[1] between headline’s sentiment of news and their impact on stock prices changes. To show this relationship, we applied the sentiment data and the price difference between the day before the news was published and the day of the news, to machine learning regression and classification models. Regression is used to predict changes and classification is used to decide whether to buy or sell stocks.  We used three stock datasets named Apple, Amazon and AXP and the results are shown in the mentioned dataset that using news with negative sentiments can make predictions just as correctly as using news with both positive and negative sentiments. In regression and classification models, Random Forest outperformed other machine learning algorithms in predicting stock price changes using news sentiment analysis. Additionally, we depicted that the results of computer and human tagging were almost similar, showing that using computer tools for text tagging will allow to tag text much more quickly and easily.

This paper investigated the optimization, modelling and effect of welding parameters on the tensile shear load bearing capacity of double pulse resistance spot welded DP590 steel. Optimization of  welding parameters was performed using the Taguchi design of experiment method. A relationship between input welding paramaters i.e., second pulse welding current, second pulse welding current time and first pulse holding time and output response i.e, tensile shear peak load was established using regression and neural network. Results showed that maximum average tensile shear peak load of 26.47 was achieved at optimum welding parameters i.e., second pulse welding current of 7.5 kA, second pulse welding time of 560 ms and first pulse holding time of 400 ms. It was also found that the ANN model predicted the tensile shear load with higher accuracy than the regression model.

Pavement deterioration is a factor contributing to the higher roughness of the road and lower driving comfort in road trips. The pavement roughness measurement has long been an especially important topic to the practitioners. The pavement condition index (PCI) and the international roughness index (IRI) have been applied as two key measures of the quality and performance of the road pavement at the level of project or the road network. Since the pavement condition assessment across a road network by means of the PCI is highly expensive, the present research aims at estimating the PCI based on the IRI to save time and money while evaluating the pavement condition across a road network. For this purpose, we considered three IRI and PCI measurements over main suburban roads in a cold-weather area during a period of 6 years. Firstly, the independence of errors was controlled using the Durbin – Watson statistic, followed by evaluating the normality of the data using the Kolmogorov – Smirnov and the Shapiro – Wilk tests. In order to measure the correlation of the normal data, we used the Pearson’s tests. Results of regression analysis between IRI and PCI showed that the best equation describing their association is a linear inverse model with an R2 = 0.992. A statistical comparison was made between the observed and predicted PCI values and model developed in the present research was further compared to similar research works, exhibiting acceptable agreement of the results.

A concept to heal the damaged surface through healing agents in metallic composites is expensive, as the selection of design parameters for such new generation smart composites are at their initial developmental stage. In the present study, the two case studies based on self-healing smart structure are included with different input design parameters to evaluate the mending of crack after healing. Taguchi based L8 experiments were conducted to analyze the influential parameters responsible for higher self-healing assessments (i.e. recovery in crack width, recovery in crack depth and flexural strength recovery). For evaluating the self-healing assessments, a soft computing technique based on S/N ratio obtained for ANOVA analysis is considered for constructing the fuzzy logic predictive model. Further, Linear regression models i.e. a statistical tool is generated to judge the accuracy of the predicted model through various error analyses. Based on S/N ratio Fuzzy logic model, results show less error values of 6.33 % and 4.94 % for case study-I and case study-II respectively in comparison to the regression model adapted for all self-healing assessments. This provides a close resemblance with the experimental observations even with less number of experimental runs. It concludes that the fuzzy logic model provides a powerful soft computing tool to perform large research work related to the design of input parameters for metallic based self-healing composites structures in near future.

The prediction of responses of the reinforced concrete shear walls subject to strong ground motions is critical in designing, assessing, and deciding the recovery strategies. This study evaluates the ability of regression models and a hybrid technique (ANN-SA model), the artificial neural network (ANN), and Simulated Annealing (SA), to predict responses of the reinforced concrete shear walls subject to strong ground motions. To this end, four buildings (15, 20, 25, and 30-story) with concrete shear walls were analyzed in OpenSees.150 seismic records are used to generate a comprehensive database of input (characteristics of records) and output (responses). The maximum acceleration, maximum velocity, and earthquake characteristics are used as predictors. Different machine learning models are used, and the accuracy of the models in identifying the responses of the shear walls is compared. The sensitivity of input variables to the seismic demand model is investigated. It has been seen from the results that the ANN-SA model has reasonable accuracy in the prediction.

The power density of a direct methanol fuel cell (DMFC) stack as a function of temperature, methanol concentration, oxygen flow rate, and methanol flow rate was studied using a response surface methodology (RSM) to maximize the power density. The operating variables investigated experimentally include temperature (50-75 °C), methanol concentration (0.5-2 M), methanol flow rate (15-30 ml min-1), and oxygen flow rate (900-1800 ml min-1). A new design of the central composite design (CCD) for a wide range of operating variables that optimize the power density was obtained using a quadratic model. The optimum conditions that yield the highest maximum power density of 86.45 mW cm-2 were provided using 3-cell stack at a fuel cell temperature of 75 °C with a methanol flow rate of 30 ml min-1, a methanol concentration of 0.5 M, and an oxygen flow rate of 1800 ml min-1. Results showed that the power density of DMFC increased with an increase in the temperature and methanol flow rate. The experimental data were in good agreement with the model predictions, demonstrating that the regression model was useful in optimizing maximum power density from the independent operating variables of the fuel cell stack.