chromium/content/browser/preloading/preloading_prediction.h

// Copyright 2022 The Chromium Authors
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file.

#ifndef CONTENT_BROWSER_PRELOADING_PRELOADING_PREDICTION_H_
#define CONTENT_BROWSER_PRELOADING_PRELOADING_PREDICTION_H_

#include <optional>
#include <string_view>

#include "base/timer/elapsed_timer.h"
#include "content/browser/preloading/preloading_confidence.h"
#include "content/public/browser/preloading_data.h"
#include "services/metrics/public/cpp/ukm_source_id.h"
#include "url/gurl.h"

namespace content {

// PreloadingPrediction keeps track of every preloading prediction associated
// with various predictors as defined in content/public/preloading.h
// (please see for more details); whether the prediction is accurate or not;
// whether the prediction is confident enough or not.
class PreloadingPrediction { … };

// The output of many predictors is a logit/probability score. To use this score
// for binary classification, we compare it to a threshold. If the score is
// above the threshold, we classify the instance as positive; otherwise, we
// classify it as negative. Threshold choice affects classifier precision and
// recall. There is a trade-off between precision and recall. If we set the
// threshold too low, we will have high precision but low recall. If we set the
// threshold too high, we will have high recall but low precision. To choose the
// best threshold, we can use ROC curves, precision-recall curves, or
// logit-precision and logit-recall curves. `ExperimentalPreloadingPrediction`
// helps us collect the UMA data required to achieve this.
class ExperimentalPreloadingPrediction { … };

// Stores data relating to a prediction made by the preloading ML model. Once
// the outcome of whether the prediction is accurate is known, the provided
// callback is invoked.
class ModelPredictionTrainingData { … };

}  // namespace content

#endif  // CONTENT_BROWSER_PRELOADING_PRELOADING_PREDICTION_H_