Learning from multirelational data through multiple views
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
Since their first release in the 1970s, relational databases have been routinely used to collect and organize real-world data---from financial transactions, marketing surveys, to health informatics observations. Traditional data mining methods expect data in the form of a single table, thus resulting in an inability to deal with such relational repositories. Multirelational Data Mining, on the other hand, aims to discover useful patterns across multiple inter-connected relations in a relational database. To this end, this work focuses on how to build classification models for relational databases through multiple views (feature sets). This study developed four multiple view strategies for mining multirelational data. The thesis firstly introduces the Multi-View Relational Classification (MRC) framework, for constructing hypotheses from sets of attributes of the presented data. The MRC strategy distinguishes itself from existing multirelational mining algorithms by excluding the need to either transform multiple relations into a universal single table or to devise new techniques for direct relational learning. The MRC algorithm offers both predictive performance and efficiency gains over current relational models, when mining diverse relational databases. Secondly, the MRC-IM method extends the MRC approach in order to deal with skew-class multirelational data. Here, the number of examples from one class is much higher than the others and correctly classifying the underrepresented examples is of prime importance. The MRC-IM method offers performance gains over a current relational model not only against majority class instances, but also against underrepresented examples. While the MRC and MRC-IM methods construct an individual view using features within a sole relation, the third multi-view strategy formulated by this work, namely the MRC-Cross approach, enables the search and collection of relevant attributes across multiple relations when constructing individual views. Finally, we present the SESP technique for pre-pruning uninteresting relations of complex relational databases. Through identifying uninteresting views from the MRC framework, our SESP method creates a pruned structure, while minimizing predictive performance loss on the final classification model. The results of this study thus suggest that learning from multiple views sets a new direction for efficiently mining data in many relational forms, including relational databases, graphs and social networks.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.001 |
| Science and technology studies | 0.001 | 0.000 |
| Scholarly communication | 0.000 | 0.001 |
| Open science | 0.003 | 0.002 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it