gradient descent algorithm

The reliability of software product is seen as critical quality factor that cannot be overemphasized. Since real world application is loaded with high amount of uncertainty, such as applicable to software reliability, there should be a technique of dealing with such uncertainty. This paper presents a reliability model to effectively handle uncertainty in software data to enhance reliability prediction of software at the early (requirements and design) stages of Software Development Life Cycle (SDLC). In this paper, a hybrid methodology of Takagi Sugeno Kang (TSK)-based Interval Type-2 Fuzzy Logic System (IT2FLS) with Artificial Neural Network (ANN) learning is employed for the prediction of software reliability. The parameters of the model are optimized using Gradient Descent (GD) back-propagation method. Relevant reliability software requirement and design metrics and software size metrics are utilized as inputs. The proposed approach uses twenty-eight real software project data. The performance of the model is evaluated using five performance metrics and found to provide output values that are very close to the actual output showing better predictive accuracy.

A novel hybrid intelligent approach for tuning the parameters of Interval Type-2 Intuitionistic Fuzzy Logic System (IT2IFLS) is introduced for the modeling and prediction of coronavirus disease 2019 (COVID-19) time series. COVID-19 is known to be a virus caused by Severe Acute Respiratory Syndrome Coronavirus 2 (SARSCoV-2) with a huge negative impact on human, work and world economy. Globally, more than 100 million people have been infected with over two million deaths and it is not certain when the pandemic will end. Predicting the trend of the COVID-19 therefore becomes an important and challenging task. Many approaches ranging from statistical approaches to machine learning methods have been formulated and applied for the prediction of the disease. In this work, the sliding mode control learning algorithm is used to adjust the parameters of the antecedent parts of  IT2IFLS system while the gradient descent backpropagation is adopted to tune the consequent parameters in a hybrid manner. The results of the hybrid intelligent learning model are compared with results of single learning models using sliding mode control and gradient descent algorithms and found to provide good performance in terms of Root Mean Squared Error (RMSE) and Mean Absolute Error (MAE) especially in noisy environments. The type-2 hybrid model also outperforms its type-1 counterparts in the different problem instances.