Landslides could cause huge damages to properties and severe loss of lives. Landslides can be detected by analyzing the environmental data collected by wireless sensor networks (WSNs). However, environmental data are usually complex and undergo rapid changes. Thus, if landslides can be predicted, people can leave the hazardous areas earlier. A good prediction mechanism is, thus, critical. Currently, a widely-used method is Artificial Neural Networks (ANNs), which give accurate predictions and exhibit high learning ability. Through training, the ANN weight coefficients can be made precise enough such that the network works in analogy to a human brain. However, when there is an imbalanced distribution of data, an ANN will not be able to learn the pattern of the minority class; that is, the class having very few data samples. As a result, the predictions could be inaccurate. To overcome this shortcoming of ANNs, this work proposes a model switching strategy that can choose between different predictors, according to environmental states. In addition, ANN-based error models have also been designed to predict future errors from prediction models and to compensate for these errors in the prediction phase. As a result, our proposed method can improve prediction performance, and the landslide prediction system can give warnings, on average, 44.2 min prior to the occurrence of a landslide.

In systems connected to smart grids, smart meters with fast and efficient responses are very helpful in detecting anomalies in realtime. However, sending data with a frequency of a minute or less is not normal with today’s technology because of the bottleneck of the communication network and storage media. Because mitigation cannot be done in realtime, we propose prediction techniques using Deep Neural Network (DNN), Support Vector Regression (SVR), and k-Nearest Neighbors (KNN). In addition to these techniques, the prediction timestep is chosen per day and wrapped in sliding windows, and clustering using Kmeans and intersection Kmeans and HDBSCAN is also evaluated. The predictive ability applied here is to predict whether anomalies in electricity usage will occur in the next few weeks. The aim is to give the user time to check their usage and from the utility side, whether it is necessary to prepare a sufficient supply. We also propose the latency reduction to counter higher latency as in the traditional centralized system by adding layer Edge Meter Data Management System (MDMS) and Cloud-MDMS as the inference and training model. Based on the experiments when running in the Raspberry Pi, the best solution is choosing DNN that has the shortest latency 1.25 ms, 159 kB persistent file size, and at 128 timesteps.

Irregularities in the use of electrical energy could result in failure of power grids and blackouts. Anomaly detection is required not only for ensuring grid safety, but also to prevent illegal hacking. Long-term data are recorded for such anomaly detection. However, due to the nonlinear characteristics of the time series data, correlation regression becomes difficult when using non-deep learning techniques. To address this issue, we use Long-Short Term Memory (LSTM) recurrency techniques. We experiment with two datasets from two different smart meters, which are sampled once every 30 seconds for one month. A total of 120,000 data samples were used for training and 40,000 data samples for testing. From the experiment results, the testing accuracy, True-Positive Rate, and False Positive Rate were 0.92, 0.81, and 0.50, respectively. Further, to demonstrate that the LSTM model can actually be designed at the network edge, we implemented the model and the trained weights on a Raspberry Pi platform. The inference time for each sample was 935 μs, which is short enough for realizing edge-based anomaly detection.

The increasing number of passengers from Jakarta to Surabaya each year will directly impact on the airport capacity in serving aviation traffic demand as it relates directly to flight frequency, runway capacity, terminal building capacity and others. The mode of Jakarta-Surabaya transportation for passengers is still focused on the mode of aircraft compared to other modes such as train. The imbalance between these modes of transportation indicates the need for other alternative modes to strengthen competitiveness in the improvement of service to passengers. This research was conducted to analyze the model of transportation mode selection between aircraft and high speed train of Jakarta-Surabaya route as an empirical case study, assuming the construction of high speed train infrastructure has been completed and ready to operate connecting Jakarta-Surabaya. This study offers the same travel time and fare between the aircraft and high speed train that reflects the special characteristics that become the passenger's main choice. Stated Preference techniques and binomial logit model are used in this study as tools for data collection and processing. As a result, the most influential passenger characteristic in this study is the level of passenger income level.

Fatigue driving is one of the main causes of traffic accidents. If we can predict that a driver will be fatigued in the near future, and give warnings and parking suggestions to let the driver stop and rest, traffic accidents can be reduced effectively. In the past, most research work on fatigue driving used conventional Back Propagation Neural Network (BPNN) model or Support Vector Machine (SVM) to make predictions. BPNN and SVM could not co-relate sequential instances of data in that series, thus making predictions not so accurate. A system is proposed that uses Recurrent Neural Network (RNN) with Long Short-term Memory (LSTM) building blocks to predict driver fatigue. Two parameters, Heart Rate Variability (HRV) and the Percentage of Eyelid Closure (PERCLOS) are used as an input data to the LSTM model so as to predict whether the driver will be fatigued in some future time slot. Experiments show that the system PERCLOS best using LSTM-based driver fatigue in prediction module that can achieve a true positive rate of 75% and an accuracy of 88%. Whereas fatigue prediction of the HRV best uses BPNN-based prediction module that yields a true positive rate and accuracy of 80% and 88% respectively.