Smartphones have been used for recognizing different transportation states. However, current studies focus on the speed of the object, which only relies on the GPS sensor rather than considering other suitable sensors and actual application factors. In this study, we propose a novel method that considers these factors comprehensively to enhance transportation state recognition. The deep Bi-LSTM (bidirectional long short-term memory) neural network structure, the crowd-sourcing model, and the TensorFlow deep learning system are used to classify the transportation states. Meanwhile, the data captured by the accelerometer and gyroscope sensors of smartphone is used to test and adjust the deep Bi-LSTM neural network model, making it easy to transfer the model into smartphones and conduct real-time recognition. The experimental results show that this study achieves transportation activity classification with an accuracy of up to 92.8%. The model of the deep Bi-LSTM neural network can be used for other time-series fields such as signal recognition and action analysis.

No abstract

No abstract

Script event prediction involves building a model using recent event sequence records to capture the correlation between events and predict future ones. However, existing methods require a large amount of training data to perform well. In many real-world scenarios, labeled data is scarce and data distribution is uneven, making it challenging to achieve satisfactory performance. Additionally, the learning process of many models demands significant computational resources, which can render them unsuitable for low-resource settings. In order to fill the task blank of event prediction in low resource scenarios, a general framework called Meta-memory low resource network learning model(MLN) is proposed. Firstly, MLN captures the memory of events that occurred in the history through trend learner, learns the context-implied representation between events, and can predict future events based on the implied representation. Secondly, a Meta-memory predictor is proposed, which can further aggregate similar contextual implicit representations to form meta-memory representations, To realize the long-term memory learning process under the condition of few-shot. Finally, a gated network combination method combines the trend learner with the Meta-memory predictor to improve the effect of event prediction. We conducted experiments on four real data sets, and the results showed the effectiveness of MLN.

No abstract