Time Series Forecasting (TSF) is key functionality in numerous fields, such as financial investment, weather services, and energy management. Although increasingly capable TSF methods occur, many of them require domain-specific data collection and model training and do not generalize well when applied in other domains. Time Series Foundation Models (TSFMs) that are pre-trained on massive heterogeneous time series data aim to overcome these limitations. The prospects for generalizability have spurred the development of a new generation of TSFMs. This study proposes a benchmark, TSFM-Bench, to facilitate comprehensive and unified evaluation of TSFMs. TSFM-Bench covers a wide range of TSFMs, including those based on large language models and those pre-trained on time series data. TSFM-Bench supports multiple forecasting scenarios, including zero-shot, few-shot, and full-shot, enabling assessment across the full range of adaptation strategies. TSFM-Bench also provides a standardized experimental protocols for critical evaluation processes such as dataset splitting, loading, normalization, and few-shot sampling, facilitating consistency and fairness. We report on an extensive evaluation of TSFMs across a diverse range of datasets spanning multiple domains and exhibiting varied statistical characteristics. Specifically, we identify pros and cons and inherent limitations of existing TSFMs, and we propose potential directions for new model designs.

Time series anomaly detection (TSAD) plays an important role in many domains such as finance, transportation, and healthcare. With the ongoing instrumentation of reality, more time series data will be available, leading also to growing demands for TSAD. While many TSAD methods already exist, new and better methods are still desirable. However, effective progress hinges on the availability of reliable means of evaluating new methods and comparing them with existing methods. We address deficiencies in current evaluation procedures related to datasets and experimental settings and protocols. Specifically, we propose a new time series anomaly detection benchmark, called TAB. First, TAB encompasses 29 public multivariate datasets and 1,635 univariate time series from different domains to facilitate more comprehensive evaluations on diverse datasets. Second, TAB covers a variety of TSAD methods, including Non-learning, Machine learning, Deep learning, LLM-based, and Time-series pre-trained methods. Third, TAB features a unified and automated evaluation pipeline that enables fair and easy evaluation of TSAD methods. Finally, we employ TAB to evaluate existing TSAD methods and report on the outcomes, thereby offering a deeper insight into the performance of these methods.

Significant wave height (SWH) is a vital metric in marine science, and accurate SWH estimation is crucial for various applications, e.g., marine energy development, fishery, early warning systems for potential risks, etc. Traditional SWH estimation methods that are based on numerical models and physical theories are hindered by computational inefficiencies. Recently, machine learning has emerged as an appealing alternative to improve accuracy and reduce computational time. However, due to limited observational technology and high costs, the scarcity of real-world data restricts the potential of machine learning models. To overcome these limitations, we propose an ocean SWH estimation framework, namely Orca. Specifically, Orca enhances the limited spatio-temporal reasoning abilities of classic LLMs with a novel spatiotemporal aware encoding module. By segmenting the limited buoy observational data temporally, encoding the buoys' locations spatially, and designing prompt templates, Orca capitalizes on the robust generalization ability of LLMs to estimate significant wave height effectively with limited data. Experimental results on the Gulf of Mexico demonstrate that Orca achieves state-of-the-art performance in SWH estimation.

Additive Kernel SVM has been extensively used in many applications, including human activity detection and pedestrian detection. Since training an additive kernel SVM model is very time-consuming, which is not scalable to large-scale datasets, many efficient solutions have been developed in the past few years. However, most of the existing methods normally fail to achieve one of these three important conditions which are (1) low classification error, (2) low memory space, and (3) low training time. In order to simultaneously fulfill these three conditions, we develop the new piecewise-linear approximate measure (PLAME) for additive kernels. By incorporating PLAME with the well-known dual coordinate descent method, we theoretically show that this approach can achieve the above three conditions. Experimental results on twelve real datasets show that our approach can achieve the best trade-off between the accuracy, memory space, and training time compared with different types of state-of-the-art methods.

In distributed storage systems (e.g., HDFS, Amazon S3, Databricks), partitioning is applied on a dataset in order to enhance performance and availability. Recently, partitioning methods have been designed to optimize the query performance of partitions with respect to the historical query workload. Never-theless, in practice, future query workloads may deviate from the historical query workload, thus deteriorating the performance of existing partitioning methods. To fill this research gap, we model the variance of future query workloads from the historical query workload, then exploit this characteristic to produce partitions that perform well for future query workloads. In addition, we explore the space of irregular shaped partition regions to further optimize the query performance. Experimental results on TPC-H and real datasets show that our proposal is up to 70x more efficient than the state-of-the-art method.

202402 bcch

Network kernel density visualization, or NKDV, has been extensively used to visualize spatial data points in various domains, including traffic accident hotspot detection, crime hotspot detection, disease outbreak detection, and business and urban planning. Due to a wide range of applications for NKDV, some geographical software, e.g., ArcGIS, can also support this operation. However, computing NKDV is very time-consuming. Although NKDV has been used for more than a decade in different domains, existing algorithms are not scalable to million-sized datasets. To address this issue, we propose three efficient methods in this paper, namely aggregate distance augmentation (ADA), interval augmentation (IA), and hybrid augmentation (HA), which can significantly reduce the time complexity for computing NKDV. In our experiments, ADA, IA and HA can achieve at least 5x to 10x speedup, compared with the state-of-the-art solutions.

In the era of smartphones, massive data are generated with geo-related info. A large portion of them come from UGC applications (e.g., Twitter, Instagram), where the content provider are users themselves. Such applications are highly attractive for targeted marketing and recommendation, which have been well studied in recommendation system. In this paper, we consider this from a brand new spatial aspect using UGC contents only. To do this we first representing each message as a point with its geo info as its location and then grouping all the points by their keywords to form multiple point groups. We form a similarity search problem that given a query keyword, our problem aims to find k keywords with the most similar distribution of locations. Our case study shows that with similar distribution, the keywords are highly likely to have semantic connections. However, the performance of existing solutions degrades when different point groups have significant overlapping, which frequently happens in UGC contents. We propose efficient techniques to process similarity search on this kind of point groups. Experimental results on Twitter data demonstrate that our solution is faster than the state-of-the-art by up to 6 times.

Location-based social media (e.g., Twitter, Foursquare) have been generating massive amount of geo-textual data. In this paper, we represent the spatial distribution of a keyword by the group of locations tagged with such keyword. Given a query keyword, our problem is to find k keywords with the most similar distribution of locations. Such query finds applications in targeted marketing and recommendation. The performance of existing solutions degrade when different point groups have significant overlapping, which happens rather frequently in real data. We propose efficient techniques to process similarity search on point groups. Experimental results on Twitter data demonstrate that our solution is faster than the state-of-the-art by up to 6 times.