Geographically and temporally neural network weighted regression for modeling spatiotemporal non-stationary relationships
04 Mar 2021-International Journal of Geographical Information Science (Taylor & Francis)-Vol. 35, Iss: 3, pp 582-608
TL;DR: To address complex non-linear interactions between time and space, a spatiotemporal proximity neural network (STPNN) is proposed in this paper to accurately generate space-time distance and has the potential to handle complex spatiotmporal non-stationarity in various geographical processes and environmental phenomena.
Abstract: Geographically weighted regression (GWR) and geographically and temporally weighted regression (GTWR) are classic methods for estimating non-stationary relationships. Although these methods have be...
Citations
More filters
TL;DR: In this paper, the authors used geospatial artificial intelligence approaches such as self-organizing map (SOM), artificial neural network (ANN), and co-active neuro-fuzzy inference system (CANFIS) to simulate groundwater quality in the Mazandaran plain in the north of Iran.
Abstract: Water quality experiments are difficult, costly, and time-consuming. Therefore, different modeling methods can be used as an alternative for these experiments. To achieve the research objective, geospatial artificial intelligence approaches such as the self-organizing map (SOM), artificial neural network (ANN), and co-active neuro-fuzzy inference system (CANFIS) were used to simulate groundwater quality in the Mazandaran plain in the north of Iran. Geographical information system (GIS) techniques were used as a pre-processer and post-processer. Data from 85 drinking water wells was used as secondary data and were separated into two splits of (a) 70 percent for training (60% for training and 10% for cross-validation), and (b) 30 percent for the test stage. The groundwater quality index (GWQI) and the effective water quality factors (distance from industries, groundwater depth, and transmissivity of aquifer formations) were implemented as output and input variables, respectively. Statistical indices (i.e., R squared (R-sqr) and the mean squared error (MSE)) were utilized to compare the performance of three methods. The results demonstrate the high performance of the three methods in groundwater quality simulation. However, in the test stage, CANFIS (R-sqr = 0.89) had a higher performance than the SOM (R-sqr = 0.8) and ANN (R-sqr = 0.73) methods. The tested CANFIS model was used to estimate GWQI values on the area of the plain. Finally, the groundwater quality was mapped in a GIS environment associated with CANFIS simulation. The results can be used to manage groundwater quality as well as support and contribute to the sustainable development goal (SDG)-6, SDG-11, and SDG-13.
18 citations
TL;DR: In this article, the authors identify and discuss spatial properties of data that influence the performance of machine learning for spatial domains of application and discuss their advantages and disadvantages, as well as some of the best practices in handling such properties in spatial domains.
Abstract: Properties of spatially explicit data are often ignored or inadequately handled in machine learning for spatial domains of application. At the same time, resources that would identify these properties and investigate their influence and methods to handle them in machine learning applications are lagging behind. In this survey of the literature, we seek to identify and discuss spatial properties of data that influence the performance of machine learning. We review some of the best practices in handling such properties in spatial domains and discuss their advantages and disadvantages. We recognize two broad strands in this literature. In the first, the properties of spatial data are developed in the spatial observation matrix without amending the substance of the learning algorithm; in the other, spatial data properties are handled in the learning algorithm itself. While the latter have been far less explored, we argue that they offer the most promising prospects for the future of spatial machine learning.
15 citations
TL;DR: Zhang et al. as discussed by the authors established a geographically weighted deep neural network (GWDNN) model that constructs a spatially weighted neural network through geographic location data and builds a neural network via stand factors and remote sensing factors to improve the interpretability of the spatial model of DDWV.
Abstract: In natural forest ecosystems, there is often abundant down dead wood (DDW) due to wind disasters, which greatly changes the size and structure of forests. Accurately determining the DDW volume (DDWV) is crucial for sustaining forest management, predicting the dynamic changes in forest resources and assessing the risks of natural disasters or disturbances. However, existing models cannot accurately express the significant spatial nonstationarity or complexity in their spatial relationships. To this end, we established a geographically weighted deep neural network (GWDNN) model that constructs a spatially weighted neural network (SWNN) through geographic location data and builds a neural network through stand factors and remote sensing factors to improve the interpretability of the spatial model of DDWV. To verify the effectiveness of this method, using 2019 data from Liangshui National Nature Reserve, we compared model fit, predictive ability and residual spatial autocorrelation among the GWDNN model and four other spatial models: an ordinary least squares (OLS) model, a linear mixed model (LMM), a geographically weighted regression (GWR) model and a deep neural network (DNN) model. The experimental results show that the GWDNN model is far superior to the other four models according to various indicators; the coefficient of determination R2, root mean square error (RMSE), mean absolute error (MAE), Moran’s I and Z-statistic values of the GWDNN model were 0.95, 1.05, 0.77, -0.01 and -0.06, respectively. In addition, compared with the other models, the GWDNN model can more accurately depict local spatial variations and details of the DDWV in Liangshui National Nature Reserve.
8 citations
Cites methods from "Geographically and temporally neura..."
...The GGRNN model combines the kernel function of the GWR model, which is fixed as a spatial weight, with the GRNN model (Irfan et al. 2016), and the GNNWR model combines the spatially weighted neural network (SWNN) with the OLS model (Du et al. 2020, Wu et al. 2020, 2021)....
[...]
TL;DR: iST-RF can improve predictive accuracy compared to the aspatial RF approach while enhancing interpretations of the trained model’s spatio-temporal relevance for its ensemble prediction and can help balance prediction and interpretation with fidelity in a spatial data science life cycle.
Abstract: Machine learning (ML) interpretability has become increasingly crucial for identifying accurate and relevant structural relationships between spatial events and factors that explain them. Methodolo...
8 citations
TL;DR: In this article, the authors proposed a framework based on a long short-term memory (LSTM) deep learning network to generate day-ahead hourly temperature forecasts with high spatial resolution.
Abstract: The rising temperature is one of the key indicators of a warming climate, capable of causing extensive stress to biological systems as well as built structures.Ambient temperature collected at ground level can have higher variability than regional weather forecasts, which fail to capture local dynamics. There remains a clear need for accurate air temperature prediction at the suburban scale at high temporal and spatial resolutions. This research proposed a framework based on a long short-term memory (LSTM) deep learning network to generate day-ahead hourly temperature forecasts with high spatial resolution. Air temperature observations are collected at a very fine scale (~150m) along major roads of New York City (NYC) through the Internet of Things (IoT) data for 2019-2020. The network is a stacked two layer LSTM network, which is able to process the measurements from all sensor locations at the same time and is able to produce predictions for multiple future time steps simultaneously. Experiments showed that the LSTM network outperformed other traditional time series forecasting techniques, such as the persistence model, historical average, AutoRegressive Integrated Moving Average (ARIMA), and feedforward neural networks (FNN). In addition, historical weather observations are collected from in situ weather sensors (i.e., Weather Underground, WU) within the region for the past five years. Experiments were conducted to compare the performance of the LSTM network with different training datasets: 1) IoT data alone, or 2) IoT data with the historical five years of WU data. By leveraging the historical air temperature from WU, the LSTM model achieved a generally increased accuracy by being exposed to more historical patterns that might not be present in the IoT observations. Meanwhile, by using IoT observations, the spatial resolution of air temperature predictions is significantly improved.
7 citations
References
More filters
[...]
TL;DR: Deep learning is making major advances in solving problems that have resisted the best attempts of the artificial intelligence community for many years, and will have many more successes in the near future because it requires very little engineering by hand and can easily take advantage of increases in the amount of available computation and data.
Abstract: Deep learning allows computational models that are composed of multiple processing layers to learn representations of data with multiple levels of abstraction. These methods have dramatically improved the state-of-the-art in speech recognition, visual object recognition, object detection and many other domains such as drug discovery and genomics. Deep learning discovers intricate structure in large data sets by using the backpropagation algorithm to indicate how a machine should change its internal parameters that are used to compute the representation in each layer from the representation in the previous layer. Deep convolutional nets have brought about breakthroughs in processing images, video, speech and audio, whereas recurrent nets have shone light on sequential data such as text and speech.
46,982 citations
Book•
[...]
TL;DR: Deep learning as mentioned in this paper is a form of machine learning that enables computers to learn from experience and understand the world in terms of a hierarchy of concepts, and it is used in many applications such as natural language processing, speech recognition, computer vision, online recommendation systems, bioinformatics, and videogames.
Abstract: Deep learning is a form of machine learning that enables computers to learn from experience and understand the world in terms of a hierarchy of concepts. Because the computer gathers knowledge from experience, there is no need for a human computer operator to formally specify all the knowledge that the computer needs. The hierarchy of concepts allows the computer to learn complicated concepts by building them out of simpler ones; a graph of these hierarchies would be many layers deep. This book introduces a broad range of topics in deep learning. The text offers mathematical and conceptual background, covering relevant concepts in linear algebra, probability theory and information theory, numerical computation, and machine learning. It describes deep learning techniques used by practitioners in industry, including deep feedforward networks, regularization, optimization algorithms, convolutional networks, sequence modeling, and practical methodology; and it surveys such applications as natural language processing, speech recognition, computer vision, online recommendation systems, bioinformatics, and videogames. Finally, the book offers research perspectives, covering such theoretical topics as linear factor models, autoencoders, representation learning, structured probabilistic models, Monte Carlo methods, the partition function, approximate inference, and deep generative models. Deep Learning can be used by undergraduate or graduate students planning careers in either industry or research, and by software engineers who want to begin using deep learning in their products or platforms. A website offers supplementary material for both readers and instructors.
38,208 citations
Journal Article•
TL;DR: It is shown that dropout improves the performance of neural networks on supervised learning tasks in vision, speech recognition, document classification and computational biology, obtaining state-of-the-art results on many benchmark data sets.
Abstract: Deep neural nets with a large number of parameters are very powerful machine learning systems. However, overfitting is a serious problem in such networks. Large networks are also slow to use, making it difficult to deal with overfitting by combining the predictions of many different large neural nets at test time. Dropout is a technique for addressing this problem. The key idea is to randomly drop units (along with their connections) from the neural network during training. This prevents units from co-adapting too much. During training, dropout samples from an exponential number of different "thinned" networks. At test time, it is easy to approximate the effect of averaging the predictions of all these thinned networks by simply using a single unthinned network that has smaller weights. This significantly reduces overfitting and gives major improvements over other regularization methods. We show that dropout improves the performance of neural networks on supervised learning tasks in vision, speech recognition, document classification and computational biology, obtaining state-of-the-art results on many benchmark data sets.
33,597 citations
Posted Content•
TL;DR: Batch Normalization as mentioned in this paper normalizes layer inputs for each training mini-batch to reduce the internal covariate shift in deep neural networks, and achieves state-of-the-art performance on ImageNet.
Abstract: Training Deep Neural Networks is complicated by the fact that the distribution of each layer's inputs changes during training, as the parameters of the previous layers change. This slows down the training by requiring lower learning rates and careful parameter initialization, and makes it notoriously hard to train models with saturating nonlinearities. We refer to this phenomenon as internal covariate shift, and address the problem by normalizing layer inputs. Our method draws its strength from making normalization a part of the model architecture and performing the normalization for each training mini-batch. Batch Normalization allows us to use much higher learning rates and be less careful about initialization. It also acts as a regularizer, in some cases eliminating the need for Dropout. Applied to a state-of-the-art image classification model, Batch Normalization achieves the same accuracy with 14 times fewer training steps, and beats the original model by a significant margin. Using an ensemble of batch-normalized networks, we improve upon the best published result on ImageNet classification: reaching 4.9% top-5 validation error (and 4.8% test error), exceeding the accuracy of human raters.
17,184 citations
TL;DR: This historical survey compactly summarizes relevant work, much of it from the previous millennium, review deep supervised learning, unsupervised learning, reinforcement learning & evolutionary computation, and indirect search for short programs encoding deep and large networks.
14,635 citations