Descrição:
<h2>1. Overview</h2>
<p>
This dataset contains the aggregated and structured results of a large-scale benchmark
evaluating twelve single-pass stream-based active learning query strategies.
This is the experimental results dataset for the master's dissertation:
<strong>"A Quantitative and Comparative Analysis of Single-Pass Stream-Based Active Learning Query Algorithms".</strong>
</p>
<p>The experiments span:</p>
<ul>
<li><strong>82 datasets</strong></li>
<li><strong>5 machine learning models</strong></li>
<li><strong>12 stream-based query strategies</strong></li>
<li><strong>5 labeling budgets</strong>: 5%, 10%, 20%, 50%, and 100%</li>
<li><strong>20,000+ experimental runs</strong></li>
</ul>
<p>
Each row represents a single experimental configuration, defined by:
</p>
<pre>
(dataset, model, hyperparameters, query strategy, labeling budget)
</pre>
<p>
This file is designed for statistical analysis, ranking, and comparative evaluation
of strategies under constrained labeling scenarios.
</p>
<hr>
<h2>2. File Structure</h2>
<ul>
<li><strong>Granularity:</strong> One row per experimental run</li>
<li><strong>Primary metric:</strong> Final model accuracy</li>
<li><strong>Evaluation setting:</strong> Single-pass stream-based active learning</li>
</ul>
<hr>
<h2>3. Column Dictionary</h2>
<p>Below is the semantic definition of each column in the dataset.</p>
<hr>
<h3><code>dataset</code></h3>
<ul>
<li><strong>Type:</strong> String</li>
<li><strong>Description:</strong> Dataset used in the experiment.</li>
<li><strong>Scope:</strong> 82 unique datasets.</li>
<li><strong>Purpose:</strong> Enables cross-dataset robustness analysis.</li>
</ul>
<hr>
<h3><code>model_name</code></h3>
<ul>
<li><strong>Type:</strong> String</li>
<li><strong>Description:</strong> Machine learning algorithm used.</li>
<li><strong>Scope:</strong> 5 model families.</li>
<li><strong>Purpose:</strong> Allows studying model–strategy interaction.</li>
</ul>
<hr>
<h3><code>model_params</code></h3>
<ul>
<li><strong>Type:</strong> String (serialized dictionary)</li>
<li><strong>Description:</strong> Hyperparameters used for the model.</li>
<li><strong>Example:</strong></li>
</ul>
<pre>
{'C': 0.01}
</pre>
<ul>
<li><strong>Recommendation:</strong> Parse into dictionary for reproducibility or hyperparameter grouping.</li>
</ul>
<hr>
<h3><code>query_strategy</code></h3>
<ul>
<li><strong>Type:</strong> String</li>
<li><strong>Description:</strong> Active learning strategy used in the stream.</li>
<li><strong>Scope:</strong> 12 strategies.</li>
<li><strong>Purpose:</strong> Main variable of interest for comparative evaluation.</li>
</ul>
<hr>
<h3><code>budget</code></h3>
<ul>
<li><strong>Type:</strong> Float</li>
<li><strong>Values:</strong></li>
<ul>
<li>0.05</li>
<li>0.10</li>
<li>0.20</li>
<li>0.50</li>
<li>1.00</li>
</ul>
<li><strong>Description:</strong> Fraction of instances allowed to be labeled.</li>
<li><strong>Interpretation:</strong> Controls labeling cost.</li>
</ul>
<hr>
<h3><code>initial_score</code></h3>
<ul>
<li><strong>Type:</strong> Float</li>
<li><strong>Description:</strong> Baseline performance before applying active learning.</li>
<li><strong>Purpose:</strong> Reference point for measuring improvement.</li>
</ul>
<hr>
<h3><code>percentage_queried</code></h3>
<ul>
<li><strong>Type:</strong> Float</li>
<li><strong>Description:</strong> Actual fraction of instances labeled.</li>
<li><strong>Note:</strong></li>
<ul>
<li>May slightly differ from the defined budget due to stream dynamics.</li>
<li>Reflects real labeling consumption.</li>
</ul>
</ul>
<hr>
<h3><code>final_accuracy</code></h3>
<ul>
<li><strong>Type:</strong> Float</li>
<li><strong>Description:</strong> Final model performance after active learning.</li>
<li><strong>Metric:</strong> Classification accuracy.</li>
<li><strong>Primary evaluation metric.</strong></li>
</ul>
<hr>
<h2>4. Summary</h2>
<p>
<code>experiment_results.csv</code> is a large-scale benchmark dataset for evaluating
stream-based active learning strategies under varying labeling budgets.
</p>
<p>It supports:</p>
<ul>
<li>Cross-dataset comparisons</li>
<li>Strategy ranking</li>
<li>Budget sensitivity analysis</li>
<li>Model–strategy interaction studies</li>
<li>Efficiency and robustness evaluation</li>
</ul>
<p>
The structure is analysis-ready and designed for statistical benchmarking
and research publication purposes.
</p>