It is an era of Internet of Things, where various types of sensors, especially wireless, are widely used to collect huge amount of data to feed various systems such as surveillance, environmental monitoring, and disaster management. In these systems, wireless sensors are deployed to make decisions or to predict an event in a real-time basis. However, the accuracy of such decisions or predictions depends upon the reliability of the sensor data. Unfortunately, erroneous data are received from the sensors. Consequently, it hampers the appropriate operations of the mentioned systems, especially in making decisions and prediction. Therefore, the detection of anomaly that exists with the sensor data drew significant attention and hence, it needs to be filtered before feeding a system to increase its reliability in making decisions or prediction. There exists various sensor anomaly detection algorithms, but few of them are able to address the uncertain phenomenon, associated with the sensor data. If these uncertain phenomena cannot be addressed by the algorithms, the filtered data into the system will not be able to increase the reliability of the decision-making process. These uncertainties may be due to the incompleteness, ignorance, vagueness, imprecision and ambiguity. Therefore, in this paper we propose a new belief-rule-based association rule (BRBAR) with the ability to handle the various types of uncertainties as mentioned.The reliability of this novel algorithm has been compared with other existing anomaly detection algorithms such as Gaussian, binary association rule and fuzzy association rule by using sensor data from various domains such as rainfall, temperature and cancer cell data. Receiver operating characteristic curves are used for comparing the performance of our proposed BRBAR with the aforementioned algorithms. The comparisons demonstrate that BRBAR is more accurate and reliable in detecting anomalies from sensor data under uncertainty. Hence, the use of such algorithm to feed the decision-making systems could be beneficial. Therefore, we have used this algorithm to feed appropriate sensor data to our recently developed belief-rule-based expert system to predict flooding in an area. Consequently, the reliability and the accuracy of the flood prediction system increase significantly. Such novel algorithm (BRBAR) can be used in other areas of applications.

Sensor data are gaining increasing global attention due to the advent of Internet of Things (IoT). Reasoning is applied on such sensor data in order to compute prediction. Generating a health warning that is based on prediction of atmospheric pollution, planning timely evacuation of people from vulnerable areas with respect to prediction of natural disasters, etc., are the use cases of sensor data stream where prediction is vital to protect people and assets. Thus, prediction accuracy is of paramount importance to take preventive steps and avert any untoward situation. Uncertainties of sensor data is a severe factor which hampers prediction accuracy. Belief Rule Based Expert System (BRBES), a knowledge-driven approach, is a widely employed prediction algorithm to deal with such uncertainties based on knowledge base and inference engine. In connection with handling uncertainties, it offers higher accuracy than other such knowledge-driven techniques, e.g., fuzzy logic and Bayesian probability theory. Contrarily, Deep Learning is a data-driven technique, which constitutes a part of Artificial Intelligence (AI). By applying analytics on huge amount of data, Deep Learning learns the hidden representation of data. Thus, Deep Learning can infer prediction by reasoning over available data, such as historical data and sensor data streams. Combined application of BRBES and Deep Learning can compute prediction with improved accuracy by addressing sensor data uncertainties while utilizing its discovered data pattern. Hence, this paper proposes a novel predictive model that is based on the integrated approach of BRBES and Deep Learning. The uniqueness of this model lies in the development of a mathematical model to combine Deep Learning with BRBES and capture the nonlinear dependencies among the relevant variables. We optimized BRBES further by applying parameter and structure optimization on it. Air pollution prediction has been taken as use case of our proposed combined approach. This model has been evaluated against two different datasets. One dataset contains synthetic images with a corresponding label of PM_2.5 concentrations. The other one contains real images, PM_2.5 concentrations, and numerical weather data of Shanghai, China. We also distinguished a hazy image between polluted air and fog through our proposed model. Our approach has outperformed only BRBES and only Deep Learning in terms of prediction accuracy.

Recent technological advancements in the area of the Internet of Things (IoT) and cloud services, enable the generation of large amounts of raw data. However, the accurate prediction by using this data is considered as challenging for machine learning methods. Deep Learning (DL) methods are widely used to process large amounts of data because they need less preprocessing than traditional machine learning methods. Various types of uncertainty associated with large amounts of raw data hinder the prediction accuracy. Belief Rule-Based Expert Systems (BRBES) are widely used to handle uncertain data. However, due to their incapability of integrating associative memory within the inference procedures, they demonstrate poor accuracy of prediction when large amounts of data is considered. Therefore, we propose the integration of an associative memory based DL method within the BRBES inference procedures, allowing to discover accurate data patterns and hence, the improvement of prediction under uncertainty. To demonstrate the applicability of the proposed method, which is named BRB-DL, it has been fine tuned against two datasets, one in the area of air pollution and the other in the area of power generation. The reliability of the proposed BRB-DL method, has also been compared with other DL methods such as Long-Short Term Memory and Deep Neural Network, and BRBES by taking into account of the air quality dataset from Beijing city and the power generation dataset of a combined cycle power plant. BRB-DL outperforms the above-mentioned methods in terms of prediction accuracy. For example, the Mean Square Error value of BRB-DL is 4.12 whereas for Long-Short Term Memory, Deep Neural Network, Fuzzy Deep Neural Network, Adaptive Neuro Fuzzy Inference System and BRBES it is 18.66, 28.49, 17.05, 16.37 and 38.15 for combined cycle power plant respectively, which are significantly higher.