Automatic assessment of interactive OLAP explorations

Detecting coherent explorations in SQL workloads

“…In one of our previous works [18], we proposed an approach for detecting OLAP analyses phrased in SQL, by converting SQL queries into OLAP queries and then checking if two consecutive queries are sufficiently close in terms of OLAP operations. In our more recent work, we used supervised learning to identify a set of query features allowing to characterize focus zones in OLAP explorations [5], or to identify queries that better contribute to an exploration [4]. The present work can be seen as a continuation of those previous works, since we have the same objective as [18] and use the same technique as [4].…”

“…In our more recent work, we used supervised learning to identify a set of query features allowing to characterize focus zones in OLAP explorations [5], or to identify queries that better contribute to an exploration [4]. The present work can be seen as a continuation of those previous works, since we have the same objective as [18] and use the same technique as [4]. The main differences with these previous works is that we make no assumption about the type of queries in the workload (particularly, they may not be multidimensional queries), and we have no ground truth (i.e., no human manual inspection of each query) on the workload.…”

“…For each query, we extract a set of simple features computed from the query text and its relationship with other queries in a session. The set of features is inspired from our previous work [5,4], which models OLAP queries as a set of features capturing typical OLAP navigation.…”

“…To identify explorations from a SQL workload, we use a technique first proposed in [4] to score the quality of OLAP explorations. This technique consists of characterizing a query by a set of simple features that are intrinsic to a query or that relate the query to its neighbor in the sequence.…”

“…This technique consists of characterizing a query by a set of simple features that are intrinsic to a query or that relate the query to its neighbor in the sequence. While in [4] this technique of feature extraction was used with supervised machine learning to score the quality of OLAP explorations, in the present work we use these features to partition an SQL workload into coherent explorations, investigating three different alternatives for session segmentation:…”