Mining regional co-location pattern (RCP) in urban areas can reveal local correlations among urban facilities, which can inform urban planning and development. However, existing mining methods lack considerations of user demand for temporal accessibility and user preferences for the regional structure. Therefore, the mined results may be impractical. To address these issues, we propose a novel RCP mining method with temporal accessibility and regional structure (TARS). First, we utilized travel time as a measure of temporal accessibility between instances and proposed the concept of an accessible path to develop a method to regionalize temporal accessibility. To enhance efficiency, we developed a neighbor search algorithm and two pruning strategies to accelerate regional division. Second, we propose a measure called spatial region preference (SRP) to help select the preferred RCPs. Extensive experiments conducted on synthetic and real datasets show that our proposed RCP mining algorithm can discover more practical patterns and run faster than the selected baseline algorithms.

The interactions between structured entities play important roles in a wide range of applications such as chemistry, material science, biology, and medical science. Recently, graph-based methods have been exploited to effectively predict the interactions among structured entities. However, these methods usually only focus on structural information of the entities and are incapable of fully utilizing the interaction information between the entities. In this paper, we propose a Denoising Variational Graph of Graphs Auto-encoder (DVGGA) which follows the graph of graphs framework to capture both structural information in structured entities and interaction information among structured entities. With denoising criterion, DVGGA is able to capture the information from the useful structures of the local graph and address the overfitting issue caused by redundant substructures. Extensive experiments conducted on real-world datasets show that DVGGA outperforms the state-of-the-art structured entity interaction prediction methods.

Co-location is one of the most fundamental spatial associations in spatial data. Previous anomalous co-location detections only aim to detect abnormal inter-feature co-locations on bivariate datasets, while the abnormal intra-feature co-locations are lost, which may have limits to applications. Thus, we propose a novel multi-view anomalous co-location detection framework (MACDF). Two kinds of relationships are captured by two proposed couplings, that is, intra-feature coupling and inter-feature coupling (intra-coupling and inter-coupling for short). Moreover, the neighboring concept in co-location relationships is redefined with a mixture-considered inter-coupling for point-like instances, inspired by the concept for polygonal instances, and a concreted algorithm, i.e. multi-view low-rank analysis with direction for anomalous co-location detection (MLAD-ACD), is designed to discover the anomalous co-locations, on not only bivariate datasets but also multivariate datasets. Experiments show the instantiated algorithm MLAD-ACD can detect anomalous co-locations effectively and efficiently on spatial datasets. Specifically, MLAD-ACD obtains at least a 71% higher AUC score than the state-of-the-art algorithms in an acceptable time.

Prevalent co-location pattern mining aims to find a subset of spatial features whose instances are frequently located together in geo-space. Sub-prevalent co-location pattern mining attains co-location patterns with richer spatial relationships based on star instance model instead of clique instance model. Further, discovering dominant features in a sub-prevalent co-location pattern is important to reveal the interaction relationship between spatial features and improve the application of the pattern. However, the methods for mining sub-prevalent co-location pattern with dominant feature use a distance threshold to determine the binary neighbor relationship between spatial instances, and ignore the fuzziness of neighbor relationship which can give a rise to the fuzziness of dominance relationship between spatial features. Thus, this paper presents mining fuzzy sub-prevalent co-location pattern with dominant feature by exploring spatial instance distribution based on the fuzzy theory. Specifically, a novel pattern, fuzzy sub-prevalent co-location pattern with dominant feature, is proposed by defining the fuzzy neighbor relationship between spatial instances and the fuzzy dominant relationship between spatial features. Then, an efficient algorithm to mine the proposed patterns is designed by utilizing the anti-monotonicity of fuzzy star participation index and fuzzy star dominance index to prune unpromising patterns. The experimental results show that the proposed patterns are practical and the mining algorithm is efficient.

A co-location pattern is a subset of spatial features whose instances are frequently located together in proximate areas. Mining co-location patterns can discover spatial dependencies in spatial datasets and have particular value in many applications. However, it is challengeable to discover co-location patterns from massive spatial datasets, due to the expensive computational cost. In this paper, we present a novel parallel co-location pattern mining approach. First, dividing spatial neighbor relationships into some neighbor-dependency partitions enables to perform mining task on each partition independently in parallel. Then, a column-based calculation approach is proposed to replace the time-consuming generation of table instances for calculating the prevalence of patterns. To further reduce the search space of patterns on each partition, two pruning strategies are suggested. We implement the parallel co-location pattern mining algorithm based on neighbor-dependency partition and column calculation via MapReduce, named PCPM-NDPCC. Substantial experiments are conducted on real and synthetic datasets to examine the performance of PCPM-NDPCC. Experimental results reveal that PCPM-NDPCC has a significant improvement in efficiency than baseline algorithms and shows better scalability for massive spatial data processing.

A huge volume of spatio-textual objects generated from location-based services enable a wide range of spatial keyword queries. Recently, researchers have proposed a novel query, called Spatial Pattern Matching (SPM), which uses a pattern to capture users' intention. It has been demonstrated to be useful but computationally intractable. Existing algorithms suffer from the low efficiency issue, especially on large scale datasets. To enhance the performance of SPM, in this paper we propose a novel Efficient Spatial Pattern Matching (ESPM) algorithm, which exploits the inverted linear quadtree index and computes matched node pairs and object pairs level by level in a top-down manner. In particular, it focuses on pruning unpromising nodes and node pairs at the high levels, resulting in a large number of unpromising objects and object pairs to be pruned before accessing them from disk. Our experimental results on real large datasets show that ESPM is over one order of magnitude faster than the state-of-the-art algorithm, and also uses much less I/O cost.

With recent advances in information technologies such as global position system and mobile internet, a huge volume of spatio-textual objects have been generated from location-based services, which enable a wide range of spatial keyword queries. Recently, researchers have proposed a novel query, called Spatial Pattern Matching (SPM), which uses a pattern to capture the user's intention. It has been demonstrated to be fundamental and useful for many real applications. Despite its usefulness, the SPM problem is computationally intractable. Existing algorithms suffer from the low efficiency issue, especially on large scale datasets. To enhance the performance of SPM, in this paper we propose a novel Efficient Spatial Pattern Matching (ESPM) algorithm, which exploits the inverted linear quadtree index and computes matched node pairs and object pairs level by level in a top-down manner. In particular, it focuses on pruning unpromising nodes and node pairs at the high levels, resulting in a large number of unpromising objects and object pairs to be pruned before accessing them from disk. We experimentally evaluate the performance of ESPM on real large datasets. Our results show that ESPM is over one order of magnitude faster than the state-of-the-art algorithm, and also uses much less I/O cost.

Point of Interest (POI) is the core element of check-in data. It is an effective way to represent POI by distributed representation which can encode the information of POI into a continuous vector space. In this work, we present a hybrid model that map the concepts of network representation learning (NRL) to learn the position of POI and map the concepts of nature language processing (NLP) to learn the category of POI. The results, which are in vector form, can be widely applied to various location based services (LBS) without complicated artificial feature extraction. Further, it can also improve the performance of LBS. By a range of experiments on real datasets, we demonstrate our model's capability at characterizing POI. We also verify the effectiveness of the hybrid model on POI recommendation.

A spatial colocation pattern is a group of spatial features whose instances are frequently located together in geographic space. Discovering colocations has many useful applications. For example, colocated plant species discovered from plant distribution data sets can contribute to the analysis of plant geography, phytosociology studies, and plant protection recommendations. In this paper, we study the colocation mining problem in the context of uncertain data, as the data generated from a wide range of data sources are inherently uncertain. One straightforward method to mine the prevalent colocations in a spatially uncertain data set is to simply compute the expected participation index of a candidate and decide if it exceeds a minimum prevalence threshold. Although this definition has been widely adopted, it misses important information about the confidence which can be associated with the participation index of a colocation. We propose another definition, probabilistic prevalent colocations, trying to find all the colocations that are likely to be prevalent in a randomly generated possible world. Finding probabilistic prevalent colocations (PPCs) turn out to be difficult. First, we propose pruning strategies for candidates to reduce the amount of computation of the probabilistic participation index values. Next, we design an improved dynamic programming algorithm for identifying candidates. This algorithm is suitable for parallel computation, and approximate computation. Finally, the effectiveness and efficiency of the methods proposed as well as the pruning strategies and the optimization techniques are verified by extensive experiments with “real + synthetic” spatially uncertain data sets.