Self-attention eidetic 3D-LSTM: Video prediction models for traffic flow forecasting

Article Computer Science, Information Systems

Multiple local 3D CNNs for region-based prediction in smart cities

Yibi Chen et al.

Summary: This paper introduces a deep learning-based method for region-based prediction in smart cities, utilizing multiple local 3D CNN spatial-temporal residual networks (LMST3D-ResNet) to extract various temporal dependencies for predicting future citywide activities.

INFORMATION SCIENCES (2021)

Add to Collection

Article Transportation Science & Technology

DNEAT: A novel dynamic node-edge attention network for origin-destination demand prediction

Dapeng Zhang et al.

Summary: The research focuses on accurate and reliable demand prediction for ride-hailing service platforms. A novel neural architecture, Dynamic Node-Edge Attention Network (DNEAT), is proposed to address the unique challenges of origin-destination demand prediction. The model outperforms six baseline models and is more robust to sparse demand data in real-world ride-hailing datasets from Chengdu, China, and New York City.

TRANSPORTATION RESEARCH PART C-EMERGING TECHNOLOGIES (2021)

Add to Collection

Article Computer Science, Artificial Intelligence

A temporal-aware LSTM enhanced by loss-switch mechanism for traffic flow forecasting

Huakang Lu et al.

Summary: This paper introduces a novel LSTM network enhanced by TCC blocks and LSM mechanism for short-term traffic flow forecasting at isolated points. Compared with traditional RNN or LSTM networks, the proposed network is able to capture more distinguishable temporal features and effectively counteract noise, leading to more accurate predictions.

NEUROCOMPUTING (2021)

Add to Collection

Article Computer Science, Information Systems

A3T-GCN: Attention Temporal Graph Convolutional Network for Traffic Forecasting

Jiandong Bai et al.

Summary: An A3T-GCN model was proposed to capture global temporal dynamics and spatial correlations in traffic flows. Experimental results demonstrate the effectiveness and robustness of the model in improving prediction accuracy.

ISPRS INTERNATIONAL JOURNAL OF GEO-INFORMATION (2021)

Add to Collection

Article Business, Finance

An efficient stock market prediction model using hybrid feature reduction method based on variational autoencoders and recursive feature elimination

Hakan Gunduz

Summary: In this study, hourly directions of eight banking stocks in Borsa Istanbul were predicted using linear-based, deep-learning (LSTM) and ensemble learning (LightGBM) models. Models were trained with different feature sets and evaluated for accuracy and F-measure metrics, with results showing that incorporating features from other stocks as inputs led to better performance.

FINANCIAL INNOVATION (2021)

Add to Collection

Article Computer Science, Artificial Intelligence