Traffic flow forecasting is essential for managing congestion, improving safety, and optimizing various transportation systems. However, it remains a prevailing challenge due to the stochastic nature of urban traffic and environmental factors. Better predictions require models capable of accommodating the traffic variability influenced by multiple dynamic and complex interdependent factors. In this work, we propose a Graph Neural Network (GNN) framework to address the stochasticity by leveraging adaptive adjacency matrices using log-normal distributions and Coefficient of Variation (CV) values to reflect real-world travel time variability. Additionally, weather factors such as temperature, wind speed, and precipitation adjust edge weights and enable GNN to capture evolving spatio-temporal dependencies across traffic stations. This enhancement over the static adjacency matrix allows the model to adapt effectively to traffic stochasticity and changing environmental conditions. Furthermore, we utilize the Adaptive Conformal Prediction (ACP) framework to provide reliable uncertainty quantification, achieving target coverage while maintaining acceptable prediction intervals. Experimental results demonstrate that the proposed model, in comparison with baseline methods, showed better prediction accuracy and uncertainty bounds. We, then, validate this method by constructing traffic scenarios in SUMO and applying Monte-Carlo simulation to derive a travel time distribution for a Vehicle Under Test (VUT) to reflect real-world variability. The simulated mean travel time of the VUT falls within the intervals defined by INRIX historical data, verifying the model’s robustness.

Learned database indexes use machine learning algorithms to directly predict the location of a query key within a sorted key array, thereby eliminating index search overhead. Prior work shows learned indexes outperforming traditional indexes in both space and query time for one-dimensional data, especially for read-intensive workloads. Usually, learned indexes linearize the multi-dimensional data with a space-filling curve before generating the index.

Due to the meteoric rise in the use of location-aware systems (for ex. social media) that generate streaming spatio-textual data in a fast-paced and highly dynamic nature, faster query processing is paramount for applications that demand performance. In this paper, we present methods to efficiently process multi-keyword spatial queries with advanced conjunctive boolean constraints which include a negative keyword list on streaming data. Modern applications require quick system responses for complex queries with spatial and multiple keyword filters. Processing such large amounts of geo-tagged stream data with various keyword filters is time-consuming and inefficient with current methods. To solve this problem we build techniques for indexing, updating, and query processing that proactively prunes the search space called the PNLR-Tree based on an R-Tree. Specifically, given positive and negative keyword lists along with spatial constraints we develop techniques for enabling advanced boolean constraints like ‘But-Not’, ‘XOR’, and others on queries such as range query, top-k, kNN, and their extensions over a sliding window. Our experiments on real datasets using various real and synthetic query workloads verify that our method has increased performance over alternate techniques.

Selectivity and cardinality estimation are main driving factors for developing cheap query plans and ultimately faster query processing. Traditionally, database systems use estimation data structures, e.g., histograms, to maintain data summaries. Machine learning models have recently been employed, acting as black boxes, in several database tasks, including cardinality estimation. In the dynamic streaming environments, both estimation data structures and machine learning models struggle with adaptation for dynamic changes in data and query workloads. This paper proposes LATEST; a system module that uses machine learning to enable dynamic adaptation of estimation data structures. For spatial-keyword queries in a streaming environment, it shows on par or better performance than the state-of-the-art estimators. LATEST builds an incremental supervised learning model over a moving time window that helps the underlying system to switch among several estimation structures to keep estimation accuracy high at all times. As an incremental learner, LATEST effectively adapts to dynamic changes of both data and queries in streaming environments. Our extensive experiments on three real datasets and various query workloads verify the effectiveness of LATEST with higher accuracy and lower response times over the state-of-the-art estimators.