Defu Zhang

Bio: Defu Zhang is an academic researcher from Xiamen University. The author has contributed to research in topics: Workflow & Cloud computing. The author has an hindex of 3, co-authored 8 publications receiving 32 citations.

A New Hybrid Convolutional Neural Network and eXtreme Gradient Boosting Classifier for Recognizing Handwritten Ethiopian Characters

Abstract: Handwritten character recognition has been profoundly studied for many years in the field of pattern recognition. Due to its vast practical applications and financial implications, handwritten character recognition is still an important research area. In this research, the Handwritten Ethiopian Character Recognition (HECR) dataset has been prepared to train the model. The images in the HECR dataset were organized with more than one color pen RGB main spaces that have been size normalized to 28 × 28 pixels. The dataset is a combination of scripts (Fidel in Ethiopia), numerical representations, punctuations, tonal symbols, combining symbols, and special characters. These scripts have been used to write ancient histories, science, and arts of Ethiopia and Eritrea. In this study, a hybrid model of two super classifiers: Convolutional Neural Network (CNN) and eXtreme Gradient Boosting (XGBoost) is proposed for classification. In this integrated model, CNN works as a trainable automatic feature extractor from the raw images and XGBoost takes the extracted features as an input for recognition and classification. The output error rates of the hybrid model and CNN with a fully connected layer are compared. A 0.4630 and 0.1612 error rates are achieved in classifying the handwritten testing dataset images, respectively. Thus XGBoost as a classifier performs a better result than the traditional fully connected layer.

Exploration of Secured Workflow Scheduling Models in Cloud Environment: A Survey

Abstract: Cloud computing (CC) is a useful tool for executing complex applications. As a result of this, it has become so popular and used in diverse domains such as science, engineering, medicine. etc. CC structure is composed of a number of virtual machines(VMs) provisioned on demand and charged on a "Pay-as-you-go" basis, it is deployed in different form of access levels. Complex applications needed to be executed on clouds are represented as workflows. Workflow scheduling (WS) is one of the most important concepts in cloud computing. WS model contributes to minimizing cost, makespan and energy as well as maximize the quality of service(QoS) of applications in clouds. Despite the security constraints set by each provider, CC has become so critical due to the considerations of applications with sensitive intermediate data, this thereby requires a security level known as Secured workflow Scheduling(SWS). This security is on the level of executing workflows. It indicates that applications with sensitive interdependent data have to be protected during their execution across different cloud VMs. The addition of security in workflow execution generates time overhead, making it complex to meet up with the QoS required by the users. Some research works have proposed algorithms for providing the QoS requirements and security at the same time. In this work, we survey some existing works, by defining the factors needed in securing workflows during execution, clarifying the domains for security, sources of security threats and their solutions as well as cloud computing services that needs security and lastly classify the proposed algorithm depending cloud computing components.

Mining of Nutritional Ingredients in Food for Disease Analysis

Abstract: Suitable nutritional diets have been widely recognized as important measures to prevent and control non-communicable diseases (NCDs). However, there is little research on nutritional ingredients in food now, which are beneficial to the rehabilitation of NCDs. In this paper, we profoundly analyzed the relationship between nutritional ingredients and diseases by using data mining methods. First, more than 7000 diseases were obtained, and we collected the recommended food and taboo food for each disease. Then, referring to the China Food Nutrition , we used noise intensity and information entropy to find out which nutritional ingredients can exert positive effects on diseases. Finally, we proposed an improved algorithm named CVNDA_Red based on rough sets to select the corresponding core ingredients from the positive nutritional ingredients. To the best of our knowledge, this is the first study to discuss the relationship between nutritional ingredients in food and diseases through data mining based on rough set theory in China. The experiments on real-life data show that our method based on data mining improves the performance compared with the traditional statistical approach, with the precision of 1.682. In addition, for some common diseases, such as diabetes, hypertension and heart disease, our work is able to identify correctly the first two or three nutritional ingredients in food that can benefit the rehabilitation of those diseases. These experimental results demonstrate the effectiveness of applying data mining in selecting of nutritional ingredients in food for disease analysis.

Decomposition Based Multi-objective Workflow Scheduling for Cloud Environments

Abstract: Workflow is a group of tasks that are processed in a particular order to complete an application. Also, it is a popular paradigm used to model complex applications. Executing complex applications in a distributed system such as cloud computing implicates optimization of several conflicting objectives such as monetary cost, energy consumption, total execution time of the application (makespan), etc. Regardless of this trend, most of the workflow scheduling approaches focused on single or bi-objectives optimization problem. In this paper, we considered the problem of workflow scheduling in a cloud environment as a multi-objective optimization problem, and hence proposed a multi-objective workflow-scheduling algorithm based on decomposition (WSABD). The proposed algorithm is capable of finding optimal solutions with a single run. Our evaluation results show that, by a single run, the proposed approach manages to obtain the Pareto Front solutions which are at least as good as schedules produced by running a single-objective scheduling algorithm with contraint for multiple times.

Internet of Things Based on Electronic and Mobile Health Systems for Blood Glucose Continuous Monitoring and Management

Abstract: Over 425 million people suffer from diabetes worldwide and this number is expected to increase over the years. Rigorous and extensive research has led to the development of increasingly advanced technologies, such as continuous glucose monitoring and glucose flash monitoring. These new technologies are more promising and efficient with respect to calculating the glycemic index and are more easier to use than the glucometer technology already established in the market. However, market solutions are often highly restrictive due to their costs. In an effort to address this challenge, this article describes the Freestyle Free sensor and the associated advantages of an integrated and low-cost environment that it offers patients. The proposed environment allows continuously monitoring the blood glucose rate and provides doctors and caregivers information remotely. Additionally, the data generated will allow the application of data mining techniques in efforts aimed at understanding the disease better. The integration between the patient and the integrated environment occurs through the near-field communication sensor over an Internet of Things card, which sends the data collected for the LibreMonitor mobile application. To evaluate the integrated environment, we compared the glucose rates measured with an official Freestyle Libre software during the same period. Based on the positive results, we propose that the integrated environment is a low-cost alternative for continuous glucose monitoring of patients with diabetes.

Traffic Density Classification Using Sound Datasets: An Empirical Study on Traffic Flow at Asymmetric Roads

Abstract: Recently, with the rapid growth of Deep Learning models for solving complicated classification problems, urban sound classification techniques have been attracted more attention. In this paper, we take an investigation on how to apply this approach for the transportation domain. Specifically, traffic density classification based on the road sound datasets, which have been recorded and preprocessed on the urban road network, is taken into account. In particular, state-of-the-art methods for analyzing and extracting sound datasets have taken into account for the classification problem of traffic flow. Consequently, this study focuses on three main processes which are: i) generating image representation for the sequences of the road sound datasets; ii) proposing a convolutional neural network model for the feature extraction; iii) adopting a hybrid approach for the classification stage by combining convolutional neural network with other machine learning models. Regarding the experiment, the road sound dataset has been collected at an urban asymmetric road with different time periods (e.g., morning and evening) in order to evaluate our proposed method. Specifically, the implementations show promising results in which the accuracies are able to achieve from 92% to 95% for classifying traffic densities with different time periods.