# 2. AI-Powered Score Prediction

User Type: **Student**
Source: *Mi Digital Academy - Education CRM Features Document*

---

## 2. AI-Powered Score Prediction

### 2.1 Score Prediction Engine
**What it does:** Predicts the Student's likely score for each target exam using a machine-learning model trained on the Student's historical data: performance in online exams, practice test scores, video completion rates, time spent per topic, assessment accuracy trends, study consistency, and current mastery. The prediction updates in real time as the Student progresses and includes a confidence interval.

**Sub-features:**
- Predicted score for each target exam
- Prediction based on historical exam performance and practice scores
- Factors in video completion rates and time spent per topic
- Factors in assessment accuracy trends and study consistency
- Factors in current mastery level of topics
- Considers the difficulty level of the target exam
- Factors in the time remaining until the exam
- Prediction updates in real time as the Student progresses
- Confidence interval for the prediction
- Prediction event logging (computed, updated)
- Audit logging of the score prediction engine

**Student User Journey:**
1. Student opens a target exam and sees the predicted score (72%, confidence 68–78%).
2. The prediction reflects the Student's recent exam performance and practice scores.
3. As Student completes more practice, the prediction updates in real time.
4. The prediction accounts for the exam's difficulty and the time remaining.
5. The confidence interval narrows as more data is available.
6. Student opens Profile → "Activity" and confirms the prediction events are recorded.

**Rules & Edge Cases:**
- The prediction is computed from the Student's historical data and current mastery.
- The prediction updates in real time as the Student progresses.
- The confidence interval reflects the amount and quality of available data.
- The prediction considers the exam's difficulty and the time remaining.
- Prediction events (computed, updated) are logged with the exam and the timestamp.
- The score prediction engine is audit-logged with the account, the exam, and the timestamp.

### 2.2 Prediction Dashboard
**What it does:** Shows the Student a prediction dashboard for each target exam: the current predicted score with confidence range, a gap analysis (predicted vs. target), a subject-wise score breakdown, and a trend graph of the prediction over time. The dashboard makes the prediction actionable.

**Sub-features:**
- Current predicted score with confidence range
- Gap analysis (predicted vs. target score)
- Subject-wise score breakdown
- Trend graph (prediction improvement over time)
- "What-if" scenarios (e.g., "If I improve Math by 10%…")
- Dashboard per target exam
- Dashboard available on web and mobile
- Dashboard event logging (viewed, scenario run)
- Audit logging of the prediction dashboard

**Student User Journey:**
1. Student opens the prediction dashboard for the Board exam.
2. The predicted score (72%, 68–78%) and the gap to the target (80%) are shown.
3. The subject-wise breakdown shows Mathematics at 65% and Physics at 80%.
4. The trend graph shows the prediction improving from 60% to 72% over a month.
5. Student runs a "what-if" scenario: "If I improve Math by 10%…" and sees the new prediction.
6. Student opens the dashboard on the mobile app.
7. Student opens Profile → "Activity" and confirms the dashboard events are recorded.

**Rules & Edge Cases:**
- The gap analysis compares the predicted score with the target score.
- The subject-wise breakdown shows the predicted contribution per subject.
- The trend graph plots the prediction over time.
- A "what-if" scenario shows the effect of a hypothetical improvement.
- Dashboard events (viewed, scenario run) are logged with the exam and the timestamp.
- The prediction dashboard is audit-logged with the account, the exam, and the timestamp.

### 2.3 Prediction Accuracy and Feedback
**What it does:** Tracks how accurate the predictions have been by comparing predicted scores with actual results after exams. The Student can see the prediction accuracy history, and the platform uses the outcomes to refine future predictions. The Student receives feedback on what drove the prediction.

**Sub-features:**
- Compare predicted score with actual result after an exam
- Prediction accuracy history per exam
- Feedback on what drove the prediction (key factors)
- Platform refines predictions using actual outcomes
- Accuracy trend over time
- "Prediction was high/low by X" indicator
- Accuracy event logging (actual recorded, accuracy computed)
- Audit logging of the prediction accuracy and feedback

**Student User Journey:**
1. After the Board exam, Student records the actual result (78%).
2. The platform compares it with the prediction (72%) and shows "Prediction was low by 6%".
3. The prediction accuracy history shows the accuracy across past exams.
4. The feedback lists the key factors that drove the prediction.
5. The accuracy trend shows the predictions getting closer over time.
6. The platform refines future predictions using the outcome.
7. Student opens Profile → "Activity" and confirms the accuracy events are recorded.

**Rules & Edge Cases:**
- The accuracy is computed as the difference between the predicted and actual score.
- The accuracy history is kept per exam.
- Actual outcomes are used to refine future predictions.
- The "high/low by X" indicator shows the direction and size of the error.
- Accuracy events (actual recorded, accuracy computed) are logged with the exam and the timestamp.
- The prediction accuracy and feedback is audit-logged with the account, the exam, and the timestamp.