Despite the fact that Dynamic Bayesian Network models have become a popular modelling platform to many researchers in recent years, not many have ventured into the realms of data anomaly and its implications on DBN models. An abnormal change in the value of a hidden state of a DBN will cause a ripple-like effect on all descendent states in current and consecutive slices. Such a change could affect the outcomes expected of such models. In this paper we propose a method that will detect anomalous data of past states using a trained network and data of the current network slice. We will build a model of pilot actions during a flight, this model is trained using simulator data of similar flights. Then our algorithm is implemented to detect pilot errors in the past given only current actions and instruments data.

In the last decade Dynamic Bayesian Networks (DBNs) have become one type of the most attractive probabilistic modelling framework extensions of Bayesian Networks (BNs) for working under uncertainties from a temporal perspective. Despite this popularity not many researchers have attempted to study the use of these networks in anomaly detection or the implications of data anomalies on the outcome of such models. An abnormal change in the modelled environment's data at a given time, will cause a trailing chain effect on data of all related environment variables in current and consecutive time slices. Albeit this effect fades with time, it still can have an ill effect on the outcome of such models. In this paper we propose an algorithm for pilot error detection, using DBNs as the modelling framework for learning and detecting anomalous data. We base our experiments on the actions of an aircraft pilot, and a flight simulator is created for running the experiments. The proposed anomaly detection algorithm has achieved good results in detecting pilot errors and effects on the whole system.

No abstract

Real-time walking behavior monitoring is essential in ensuring safety and improving people's physical conditions with mobility difficulties. In this paper, a real-time walking motion detection system based on the intelligent walking stick, mobile phone and multi-label imbalance classification method combining focal loss and LightGBM (MFGBoost) is proposed. The Internet of Things (IoT) technology is utilized for communicating between the walking stick and mobile phone. The new MFGBoost is embedded into the Raspberry Pi to classify human motions. MFGBoost is scalable, and other boosting models, such as XGBoost, could also be used as its base classifier. An improved derivation method of the multi-classification focal loss function is proposed in this paper, which is the key to the combination of multi-classification focal loss and Boosting algorithms. We propose a novel denoise method based on window matrix and COPOD algorithm (W-OD). The window matrix is designed to extract data features and smooth noise, and COPOD could output the noise level of the model. A weighted loss function is designed to adjust the model's attention to different samples based on the W-OD algorithm. We evaluate the latest classification model from multiple perspectives on multiple benchmark datasets and demonstrate that MFGBoost and W-OD-MFGBoost could improve classification performance and decision-making efficiency. Experiments conducted on human motion datasets show that W-OD-MFGBoost could achieve more than 97 percent classification accuracy.

This paper focuses on road surface real-time detection by using tripod dolly equipped with Raspberry Pi 3 B+, MPU 9250, which is convenient to collect road surface data and realize real-time road surface detection. Firstly, six kinds of road surfaces data are collected by utilizing Raspberry Pi 3 B+ and MPU 9250. Secondly, the classifiers can be obtained by adopting several machine learning algorithms, recurrent neural networks (RNN) and long short-term memory (LSTM) neural networks. Among the machine learning classifiers, gradient boosting decision tree has the highest accuracy rate of 97.92%, which improves by 29.52% compared with KNN with the lowest accuracy rate of 75.60%. The accuracy rate of LSTM neural networks is 95.31%, which improves by 2.79% compared with RNN with the accuracy rate of 92.52%. Finally, the classifiers are embedded into the Raspberry Pi to detect the road surface in real time, and the detection time is about one second. This road surface detection system could be used in wheeled robot-car and guiding the robot-car to move smoothly.