Showing papers by "Chittagong University of Engineering & Technology published in 2019"

Effectiveness analysis of machine learning classification models for predicting personalized context-aware smartphone usage

Abstract: Due to the increasing popularity of recent advanced features and context-awareness in smart mobile phones, the contextual data relevant to users’ diverse activities with their phones are recorded through the device logs. Modeling and predicting individual’s smartphone usage based on contexts, such as temporal, spatial, or social information, can be used to build various context-aware personalized systems. In order to intelligently assist them, a machine learning classifier based usage prediction model for individual users’ is the key. Thus, we aim to analyze the effectiveness of various machine learning classification models for predicting personalized usage utilizing individual’s phone log data. In our context-aware analysis, we first employ ten classic and well-known machine learning classification techniques, such as ZeroR, Naive Bayes, Decision Tree, Random Forest, Support Vector Machine, K-Nearest Neighbors, Adaptive Boosting, Repeated Incremental Pruning to Produce Error Reduction, Ripple Down Rule Learner, and Logistic Regression classifiers. We also present the empirical evaluations of Artificial Neural Network based classification model, which is frequently used in deep learning and make comparative analysis in our context-aware study. The effectiveness of these classifier based context-aware models is examined by conducting a range of experiments on the real mobile phone datasets collected from individual users. The overall experimental results and discussions can help both the researchers and applications developers to design and build intelligent context-aware systems for smartphone users.

A Color and Texture Based Approach for the Detection and Classification of Plant Leaf Disease Using KNN Classifier

Abstract: Modern organic farming is gaining popularity in the agriculture of many developing countries. There are many problems arise in farming due to various environmental factors and among these plant leaf disease is considered to be the most strong factor that causes the deficit of agricultural product quality. The goal is to mitigate this issue through computer vision and machine learning technique. This paper proposed a technique for plant leaf disease detection and classification using K-nearest neighbor (KNN) classifier. The texture features are extracted from the leaf disease images for the classification. In this work, KNN classifier will classify the diseases like alternaria alternata, anthracnose, bacterial blight, leaf spot, and canker of various plant species. The proposed approach can successfully detect and recognize the selected diseases with 96.76 % accuracy.

Context-aware rule learning from smartphone data: survey, challenges and future directions

Abstract: Smartphones are considered as one of the most essential and highly personal devices of individuals in our current world. Due to the popularity of context-aware technology and recent developments in smartphones, these devices can collect and process raw contextual data about users’ surrounding environment and their corresponding behavioral activities with their phones. Thus, smartphone data analytics and building data-driven context-aware systems have gained wide attention from both academia and industry in recent days. In order to build intelligent context-aware applications on smartphones, effectively learning a set of context-aware rules from smartphone data is the key. This requires advanced data analytical techniques with high precision and intelligent decision making strategies based on contexts. In comparison to traditional approaches, machine learning based techniques provide more effective and efficient results for smartphone data analytics and corresponding context-aware rule learning. Thus, this article first makes a survey on previous work in the area of contextual smartphone data analytics and then presents a discussion of challenges and future directions for effectively learning context-aware rules from smartphone data, in order to build rule-based automated and intelligent systems.

The use of watershed geomorphic data in flash flood susceptibility zoning: a case study of the Karnaphuli and Sangu river basins of Bangladesh

Abstract: The occurrence of heavy rainfall in the south-eastern hilly region of Bangladesh makes this area highly susceptible to recurrent flash flooding. As the region is the commercial capital of Bangladesh, these flash floods pose a significant threat to the national economy. Predicting this type of flooding is a complex task which requires a detailed understanding of the river basin characteristics. This study evaluated the susceptibility of the region to flash floods emanating from within the Karnaphuli and Sangu river basins. Twenty-two morphometric parameters were used. The occurrence and impact of flash floods within these basins are mainly associated with the volume of runoff, runoff velocity, and the surface infiltration capacity of the various watersheds. Analysis showed that major parts of the basin were susceptible to flash flooding events of a ‘moderate’-to-‘very high’ level of severity. The degree of susceptibility of ten of the watersheds was rated as ‘high’, and one was ‘very high’. The flash flood susceptibility map drawn from the analysis was used at the sub-district level to identify populated areas at risk. More than 80% of the total area of the 16 sub-districts were determined to have a ‘high’-to-‘very-high’-level flood susceptibility. The analysis noted that around 3.4 million people reside in flash flood-prone areas, therefore indicating the potential for loss of life and property. The study identified significant flash flood potential zones within a region of national importance, and exposure of the population to these events. Detailed analysis and display of flash flood susceptibility data at the sub-district level can enable the relevant organizations to improve watershed management practices and, as a consequence, alleviate future flood risk.

Performance Analysis of Machine Learning Techniques to Predict Diabetes Mellitus

Abstract: Diabetes mellitus is a common disease of human body caused by a group of metabolic disorders where the sugar levels over a prolonged period is very high. It affects different organs of the human body which thus harm a large number of the body's system, in particular the blood veins and nerves. Early prediction in such disease can be controlled and save human life. To achieve the goal, this research work mainly explores various risk factors related to this disease using machine learning techniques. Machine learning techniques provide efficient result to extract knowledge by constructing predicting models from diagnostic medical datasets collected from the diabetic patients. Extracting knowledge from such data can be useful to predict diabetic patients. In this work, we employ four popular machine learning algorithms, namely Support Vector Machine (SVM), Naive Bayes (NB), K-Nearest Neighbor (KNN) and C4.5 Decision Tree (DT), on adult population data to predict diabetic mellitus. Our experimental results show that C4.5 decision tree achieved higher accuracy compared to other machine learning techniques.

A machine learning based robust prediction model for real-life mobile phone data

Abstract: Real-life mobile phone data may contain noisy instances, which is a fundamental issue for building a prediction model with many potential negative consequences. The complexity of the inferred model may increase, may arise over-fitting problem, and thereby the overall prediction accuracy of the model may decrease. In this paper, we address these issues and present a robust prediction model for real-life mobile phone data of individual users, in order to improve the prediction accuracy of the model. In our robust model, we first effectively identify and eliminate the noisy instances from the training dataset by determining a dynamic noise threshold using naive Bayes classifier and laplace estimator, which may differ from user-to-user according to their unique behavioral patterns . After that, we employ the most popular rule-based machine learning classification technique , i.e., decision tree, on the noise-free quality dataset to build the prediction model. Experimental results on the real-life mobile phone datasets (e.g., phone call log) of individual mobile phone users, show the effectiveness of our robust model in terms of precision, recall and f-measure.

RecencyMiner: mining recency-based personalized behavior from contextual smartphone data

Abstract: Due to the advanced features in recent smartphones and context-awareness in mobile technologies, users’ diverse behavioral activities with their phones and associated contexts are recorded through the device logs. Behavioral patterns of smartphone users may vary greatly between individuals in different contexts—for example, temporal, spatial, or social contexts. However, an individual’s phone usage behavior may not be static in the real-world changing over time. The volatility of usage behavior will also vary from user-to-user. Thus, an individual’s recent behavioral patterns and corresponding machine learning rules are more likely to be interesting and significant than older ones for modeling and predicting their phone usage behavior. Based on this concept of recency, in this paper, we present an approach for mining recency-based personalized behavior, and name it “RecencyMiner” for short, utilizing individual’s contextual smartphone data, in order to build a context-aware personalized behavior prediction model. The effectiveness of RecencyMiner is examined by considering individual smartphone user’s real-life contextual datasets. The experimental results show that our proposed recency-based approach better predicts individual’s phone usage behavior than existing baseline models, by minimizing the error rate in various context-aware test cases.

Optimal PID tuning for controlling the temperature of electric furnace by genetic algorithm

Abstract: Temperature is one of the exigent parameters that needs to be controlled in today’s industries. Importantly this temperature control should be precise and fast. As the conventional controllers are not optimally tuned, the controller used for controlling the temperature of the electric furnace does not exhibit better performance. Its rise time and settling time is too large as well as it has a sizable amount of overshoot. This paper presents a Genetic Algorithm based PID controller to overcome the low precision, long rise time and settling time of the controller. In this algorithm, Integral of Absolute Error is taken as the object function for minimizing the error. Using this function, the algorithm engenders the optimum value of the gain parameters (Kp, Ki, Kd) for the PID controller. It shows better control over the conventional controllers. As the overshoot, settling time, and rise time are substantially improved, it provides sharp and prompt control over the temperature. This precise and instant control of temperature has a great impact on the food and medicine industries. As the temperature could be controlled precisely and instantly, we can avoid the change/degradation of the physical properties of the materials that are under process.

Sentiment Analysis of Bengali Texts on Online Restaurant Reviews Using Multinomial Naïve Bayes

Abstract: Recently, determining the customer impression is considered one of the prominent factors on the success of the restaurant businesses. Due to the rapid growth of digital contents related to restaurant or foods in the web, people are more inclined on reviews before going to any restaurant so the significance of customer review is inevitable. In order to selects a restaurant customer needs to check thousands of feedback’s to understand the restaurant quality or services. Therefore, classification of a significant amount of reviews into a sentimental category is required to attain meaningful insights so that the customer can choose restaurants based on their preferences. This classification can be done by sentiment analysis. This paper proposes a system that can classify customer reviews into positive and negative classes based on their sentimental feedback. We have tested the proposed system with 1000 restaurant reviews text written in Bengali. The experimental result shows that the proposed the system can classify restaurant reviews with 80.48% accuracy using multinomial Naive Bayes.

Emerging Big Data Sources for Public Transport Planning: A Systematic Review on Current State of Art and Future Research Directions

Abstract: The rapid advancement of information and communication technology has brought a revolution in the domain of public transport (PT) planning alongside other areas of transport planning and operations. Of particular significance are the passively generated big data sources (e.g., smart cards, detailed vehicle location data, mobile phone traces, social media) which have started replacing the traditional surveys conducted onboard, at the stops/stations and/or at the household level for gathering insights about the behavior of the PT users. This paper presents a systematic review of the contemporary research papers related to the use of novel data sources in PT planning with particular focus on (1) assessing the usability and potential strengths and weaknesses of different emerging big data sources, (2) identifying the challenges and highlighting research gaps. Reviewed articles were categorized based on qualitative pattern matching (similarities/dissimilarities) and multiple sources of evidence analysis under three categories—use of big data in (1) travel pattern analysis, (2) PT modelling, and (3) PT performance assessment. The review revealed research gaps ranging from methodological and applied research on fusing different forms of big data as well as big data and traditional survey data; further work to validate the models and assumptions; lack of progress on developing more dynamic planning models. Findings of this study could inform transport planners and researchers about the opportunities/challenges big data bring for PT planning. Harnessing the full potential of the big data sources for PT planning can be extremely useful for cities in the developing world, where the PT landscape is changing more rapidly, but traditional forms of data are expensive to collect.

An Integrated CNN-RNN Framework to Assess Road Crack

Abstract: Road crack detection and road damage assessment are necessary to support driving safety in a route network. Several unexpected incidents (e.g. road accidents) take place all over the world due to unhealthy road infrastructure. This paper proposes a deep learning approach for road crack detection and road damage assessment which will contribute to the transport sector of a country like Bangladesh where a plethora of roads undergo the crack problem. The proposed model consists of two phases. In the first phase, the model is trained using transfer learning (VGG16) to detect the existence of crack on the road surface. In the second phase, an integrated framework, combining CNN(VGG16) and RNN(LSTM), is trained to classify the crack in one of the two categories-severe and slight. After experiments, the validation accuracies obtained by the proposed models (VGG16 and VGG16-LSTM) are respectively 99.67% and 97.66%.

Challenges for flood risk management in flood-prone Sirajganj region of Bangladesh

Abstract: Department of Civil Engineering, Chittagong University of Engineering & Technology, Chittagong, Bangladesh UNESCO-IHE Institute for Water Education, Delft, The Netherlands Institute of Water and Flood Management (IWFM), Bangladesh University of Engineering and Technology (BUET), Dhaka, Bangladesh Processing and Flood Forecasting Circle (PFFC), Bangladesh Water Development Board (BWDB), Dhaka, Bangladesh Institute of Water Modelling, Dhaka, Bangladesh

Evaluation of the elemental distribution in soil samples collected from ship-breaking areas and an adjacent island

Abstract: Ship-breaking yards in the world mostly reside along the coastal areas which possess diverse elemental contents due to the presence of heavy minerals. So to explain the heavy elemental enrichment of ship-breaking sites in terms of only anthropogenic contribution is insufficient. While studying elemental distribution of ship-breaking sites, choosing appropriate control sample is vital. To evaluate the elemental distribution in ship-breaking yards of Bangladesh and in an adjacent island (Sandwip), soil samples from each site and some heavy minerals have been analyzed for Na, K, Sc, Cr, Fe, Co, As, Sb, Cs, La, Ce, Sm, Hf and Th abundances by neutron activation analysis. For assessing elemental distribution, pollution load index (PLI), geo-accumulation index and inter-element correlation study were performed. Co-related variation of Cr and Fe (r = 0.651) and chemical compositions of heavy minerals invoke that a significant portion of elemental enrichment in ship-breaking yards seems to have mineralogical origin while the enrichments of As, Sb and Cs (mean abundances: 11.3 ± 3.9, 5.66 ± 4.97 and 10.9 ± 6.6 ppm, respectively) are solely originated from anthropogenic activities (e.g., ship-breaking). On the other hand, elemental distribution of Sandwip (which is sometimes used as control sample) seems to be unaffected by the mainland ship-breaking activities and possesses crustal origin, though a minute fractionation of heavy metals is observed within the east (PLI: 0.84 ± 0.05) and west (PLI: 0.52 ± 0.04) side of the island. For the first time, this study reveals the mineralogical contribution of heavy elements in ship-breaking site and will be decisive for choosing proper elemental abundances of control site.

Mechanical, electronic, optical, and thermodynamic properties of orthorhonmbic LiCuBiO4 crystal: a first–priciples study

Abstract: Density functional theory (DFT) based first-principles calculations using GGA+U method have been performed for the first time to investigate elastic, electronic, optical, thermodynamic properties including charge density, Fermi surface, Mulliken population analysis, and theoretical Vickers hardness of the newly synthesized LiCuBiO4 (LCBO) compound. The calculated structural parameters are in good agreement with available experimental results, which assessed the reliability of our calculations. The analysis of elastic constants indicates mechanical stability of the LCBO. The values of Poisson's and Pugh's ratios confirm the ductile nature of the LCBO. The mechanically anisotropy is found by the different anisotropy factors. The overlapping of valence and conduction bands near the Fermi level (EF) and the several bands crossing the EF reveal the metallic behaviour of the LCBO. The electronic charge density mapping and Mulliken population analysis exhibits a combination of covalent, ionic, and metallic bonding of the LCBO. The calculated Fermi surface comprised of two-dimensional topology due to the low-dispersion of O-2p and Cu-3d states, which implies the possible multi-band nature of LCBO. The analysis of thermodynamic and various optical properties suggest that LCBO can be a potential candidate for optoelectronic devices in the visible and ultraviolet energy regions and as a thermal barrier coating (TBC) material.

Feature extraction from dermoscopy images for melanoma diagnosis

Abstract: Melanoma is the deadliest type of skin cancer. It has been rising exponentially for the last few decades. If it is diagnosed and treated at its early stage, the survival rate is very high. To prevent the invasive biopsy technique, automated diagnosis of melanoma from dermoscopy images has become a hot research area for the last few decades. This paper proposes three new distinct and effective features with some existing features related to shape, size and color properties of dermoscopy images based on ABCD rule for melanoma detection. ABCD stands for Asymmetry, Border, Color, and Diameter of the skin lesion. A two-stage segmentation approach including Otsu algorithm and Chan–Vese algorithm for lesion segmentation is implemented in this paper. Dull-Razor algorithm removes the black and dark hair from the input images and artificial neural network classifier classifies the malignant and benign images based on the extracted features. Implementation result of the proposed approach achieves 98.2% overall classification accuracy with 98% sensitivity and 98.2% specificity. These promising results indicate that the proposed system is able to assist the dermatologists in early detection of melanoma.

Study of physical properties towards optimizing sintering temperature of Y-substituted Mg-Zn ferrites

Abstract: Optimum sintering temperature (Ts) plays an important role in controlling densification and growth of grains which greatly affect the magnetic and electrical properties of polycrystalline materials. Y-substituted Mg-Zn [Mg0.5Zn0.5YxFe2−xO4 (0 ≤ x ≤ 0.05)] ferrites have been prepared by using conventional standard ceramic technique and were sintered between 1100 and 1300 °C in the steps of 50 °C. Characterizations of the samples have been done by X-ray diffraction (XRD) technique, field emission scanning electron microscopy (FESEM), dielectric and permeability measurements to find the suitable Ts. XRD data have been analyzed and the results confirmed the same information regarding phase analysis at different Ts. Further characterization of the samples, sintered at 1100 and 1150 °C, were not continued due to their low density and high porosity. FESEM images indicate the change in microstructure with Ts. The decrease (increase) of ac electrical resistivity with Ts (Y content) has been observed while the dielectric constant behaves in opposite manner. Impedance spectroscopy also exhibited similar trend as ac electrical resistivity. The initial permeability revealed the wide stability zone of frequency and different optimum Ts for different compositions. The ac resistivity values for the compositions have been found higher at 1200 °C, but the compositions have highest value of bulk density and permeability at other Ts. The effect of Ts on the physical properties (bulk density, porosity and permeability) of Y-substituted Mg0.5Zn0.5YxFe2−xO4 (0 ≤ x ≤ 0.05) ferrites elucidate that the optimum Ts should be of 1250 °C for x = 0.01, 0.02 and 0.03 while the Ts should be of 1300 °C for x = 0.04 and 0.05.

AppsPred: Predicting Context-Aware Smartphone Apps using Random Forest Learning

Abstract: Due to the popularity of context-awareness in the Internet of Things (IoT) and the recent advanced features in the most popular IoT device , i.e., smartphone , modeling and predicting personalized usage behavior based on relevant contexts can be highly useful in assisting them to carry out daily routines and activities. Usage patterns of different categories smartphone apps such as social networking , communication, entertainment, or daily life services related apps usually vary greatly between individuals. People use these apps differently in different contexts, such as temporal context, spatial context, individual mood and preference, work status, Internet connectivity like Wifi status, or device related status like phone profile, battery level etc. Thus, we consider individuals’ apps usage as a multi-class context-aware problem for personalized modeling and prediction. Random forest learning is one of the most popular machine learning techniques to build a multi-class prediction model. Therefore, in this paper, we present an effective context-aware smartphone apps prediction model, and name it “AppsPred” using random forest machine learning technique that takes into account optimal number of trees based on such multi-dimensional contexts to build the resultant forest. The effectiveness of this model is examined by conducting experiments on smartphone apps usage datasets collected from individual users. The experimental results show that our AppsPred significantly outperforms other popular machine learning classification approaches like ZeroR, Naive Bayes, Decision Tree, Support Vector Machines , Logistic Regression while predicting smartphone apps in various context-aware test cases.