The vast amounts of spatio-temporal data generated by a variety of devices are best utilized in conjunction with relational, tabular data rather than being referenced separately. To enhance the efficiency of data analysis, learned models are frequently used to approximate query results by increasing responsiveness at the cost of some accuracy. Machine learning techniques exist for spatio-temporal data and tabular data separately, but it is not straightforward to represent the combined data in a unified learned model. This paper explores the challenge of learning from heterogeneous data, particularly trajectory and tabular data, and proposes approaches for representation learning using probabilistic circuits and deep neural networks.

The increasing importance of data-driven decision-making across various sectors, coupled with the need for efficient large-scale data analysis while upholding privacy, prompts our exploration. Accordingly, approaches like synthetic data generation and approximate query processing have arisen. In this study, we combined differential privacy with approximate query processing for machine learning to enhance privacy. Our approach has been implemented through differentially private generative models within an approximate query processing framework, all of which safeguard data privacy. We provide assessments of synthetic data quality concerning sensitive data and the relative error in approximate query processing utilizing synthetic data.

Approximate query processing can help reduce response time for aggregate queries in exploratory data analysis. In this study, we describe basic query transformation rules for processing approximate queries using synthetic data tables or inferential models. Based on the preliminary experimental results, we confirm that ML models can be used to provide approximate query results in response times acceptable for applications.

Huge amounts of data are ceaselessly being generated by a variety of devices, and the processing efforts for their collection and analysis grows exponentially as well. Storing them in one place and getting exact answers is almost impractical. Furthermore, computing aggregation and statistics that most exploratory data analysis would require imposes a heavy burden on networking and computing infrastructures. By adopting the edge/fog computing paradigm that has recently been developing can reduce such overheads by offloading jobs from central clouds to edge devices. We try to go one step further in this direction by approximating aggregate values and statistics for data analysis using tractable probabilistic models and optimizing network performance. This paper evaluates our preliminary result of our on-going project that was gained by fast-prototyping using Sum-Product Networks.

KIWI is a SQL-on-Hadoop system enabling batch and interactive analytics for big data. In database systems, materialized views, stored pre-computed results for queries, are one of the most commonly used techniques to improve the query processing speed. However, the key challenge in using materialized views is maintaining their freshness as base data changes. This paper introduces a new approach for accelerating OLAP query processing using query workload statistics and query column sets instead of materialized views. We present an architecture of SQL-on-Hadoop system using query column sets of original tables in database. The experimental results demonstrate that our system can provide improved performance by 1.77x on average in terms of TPC-H query processing.