Search

A common practice to capture the non-stationary characteristics of the time series data in Artificial Neural Network (ANN) is by randomly dividing the whole set of available data into training, validation and testing, i.e. the data in validation and testing are represented in the training data. Consequently, the usability of the developed model on data not represented by the training data used during the network model development process is always doubtful. In this work, we present a back-propagation neural network (BNN) model trained using one-day history data to predict soil moisture data at 1 km resolution for two future dates. Specifically, high soil moisture values were observed in the training data while the testing data were characterized by drier conditions due to minimal or no rainfall. Our model uses separate mean and standard deviation statistics values from the training and testing data, respectively, to the z-normalized data. With data pre-processed using this method, the BNN model next uses a moving window of size 4 km × 4 km to capture the spatial variability of the soil moisture throughout the 40 km × 40 km study area. The coupling of the normalization and moving window method managed to achieve average soil moisture with Root Mean Square (RMSE) of 3.67% and correlation coefficient,R2of 0.89. By only using the suggested normalization without the moving window method, the BNN model managed to achieve an average RMSE of barely 5.82% withR2 = 0.83. When comparing with the normal practice of using the same mean and standard deviation statistics of the training data in the testing data, the retrieval accuracy of the BNN model deteriorates to 8.86% withR2 = 0.32. The experiment results demonstrated that the proposed coupling method performed better in terms of both RMSE and R2for soil moisture retrieval.

Spectrum-based Fault Localization (SBFL) has been widely studied as a debugging technique to reduce time and effort in locating faulty code in software. In SBFL, execution profiles (spectra) of pass and fail test cases are analyzed with SBFL metric to rank software code according to their likeliness to be faulty. However, there are significantly more pass test cases than fail test cases in typical test suites for faulty programs. The domination of pass test cases creates imbalance test suites that have a negative impact on the performance of SBFL. This is attributed to the fact that the execution profiles of fail test cases provide the essential information on the location of faulty code. In this paper, we propose to clone the spectra of fail test cases beyond balanced test suite to improve the performance of SBFL. Our empirical study shows that the proposed cloning method significantly improves the performance of commonly used SBFL metrics. Furthermore, we attempt to identify the amount of cloning required to achieve the optimal performance for the proposed cloning method.

Financial charting software is widely used in share, commodity and foreign currency exchange markets to visualize and analyze price movements. Its quality is critical because incorrect outputs may lead to wrong analysis and trading decisions, and consequently substantial financial losses. Human visual judgment is often required to test financial charting software because of the graphical complexity of software outputs and limited knowledge of expected outputs. Such approach is labour intensive and error-prone. In this paper, we propose an automated testing technique combining metamorphic testing, assertion checking and a novel data label extraction method to eliminate human visual judgment from testing financial charting software. We used this technique to test pre-release builds of a commercial Point and Figure charting software component, and demonstrated that the proposed technique can effectively detect actual faults in the software component. Further, we discuss how the technique can be extended to test other charting software components.

Usual techniques used in the automatic test case generation and executions assume that a complete oracle will be available during the testing process. However, in some numerical programs this assumption is not valid because the oracle in form of analytical solution is either not available or too difficult to obtain. To overcome this problem, metamorphic testing proposes to generate follow-up test cases to check important properties of the output function. We study the intrinsic properties of Medial Axes for casting applications. Even though Medial Axes is a geometric function that is independent from the output function of casting simulation, we propose to use its intrinsic properties for metamorphic testing of casting simulation to alleviate the oracle problem.