Student Drop-off - MIND API

The drop-off prediction endpoints read live from the BigQuery warehouse (mlip_dropoff_prediction_synth.student_dashboard_30d) — this data is not stored in the MIND database; each request queries BigQuery directly (with a short server-side cache). All endpoints require STUDENT_PERFORMANCE.can_view permission.

How risk bands are derived. The model emits risk_level of High / Medium / Low plus a drop_off_score. The dashboard adds a Critical band as the top slice of High (drop_off_score ≥ critical cutoff, default 0.50). Because the synthetic drop_off_score tops out around 0.55, the table’s threshold column (0.7) is not used for flagging — the calibrated cutoffs are: flagged ≥ 0.15, high ≥ 0.40, critical ≥ 0.50 (all configurable).

Get Aggregate Dashboard Stats

GET /v1/console/student-dropoff/stats

Summary cards for the At-Risk Students dashboard: total flagged, and Critical / High / Medium counts, each with the change vs the previous model run.

Query parameters

faculty

string

Filter by faculty.

programme

string

Filter by programme.

year

string

Filter by programme level (e.g. 100L, 300L).

cohort

string

Filter by cohort (e.g. 202402_S1).

period

datetime

Model-run timestamp to report on. Defaults to the latest run. When a non-latest run is requested, card counts are derived from drop_off_score (basis scores) rather than the model labels (basis labels).

risk_level

string

Accepted for API symmetry. The band cards always show the full breakdown, so this does not change the aggregate counts (it scopes the at-risk list).

Example request

curl "https://staging-be.mind.miva.university/v1/console/student-dropoff/stats?faculty=Computing" \
  -H "Authorization: Bearer <access_token>"

Response

{
  "success": true,
  "data": {
    "total_students": 10000,
    "total_flagged": { "count": 3000, "pct_change": -0.2886 },
    "critical": { "count": 284, "pct_change": -0.0105 },
    "high": { "count": 716, "pct_change": -0.1848 },
    "medium": { "count": 2000, "pct_change": -0.3243 },
    "current_period": "2026-05-29T07:59:13.935658Z",
    "previous_period": "2026-05-26T07:59:13.935658Z",
    "cards_basis": "labels",
    "thresholds": {
      "model_threshold": 0.7,
      "flagged_score_cutoff": 0.15,
      "high_score_cutoff": 0.40,
      "critical_score_cutoff": 0.50,
      "labels": ["critical", "high", "medium", "low"]
    },
    "filters": { "faculty": "Computing" }
  },
  "message": null
}

Field reference

Field	Description
`total_flagged` / `critical` / `high` / `medium`	`{ count, pct_change }`. `pct_change` is a fraction (`-0.2886` = −28.86%) vs the previous run, derived from `run_history` scores. `null` when there is no prior run.
`cards_basis`	`labels` (counts from the model’s risk_level on the latest snapshot) or `scores` (score-derived, for a historical `period`).
`current_period` / `previous_period`	Model-run timestamps used for the comparison.
`thresholds`	The calibrated cutoffs in effect, plus the informational model threshold.

GET /v1/console/student-dropoff/filter-options

Distinct values for the dashboard filter controls.

Response

{
  "success": true,
  "data": {
    "faculties": ["Allied Health Sciences", "Communication & Media Studies", "Computing", "Management"],
    "programmes": ["BSc Accounting", "BSc Computer Science", "..."],
    "years": ["100L", "200L", "300L", "400L", "500L"],
    "cohorts": ["202401_S1", "202402_S1", "202402_S2", "202501_S1"],
    "periods": ["2026-05-29T07:59:13.935658Z", "2026-05-26T07:59:13.935658Z"]
  },
  "message": null
}

Predicted Drop-off Over Time

GET /v1/console/student-dropoff/trend

Time series of flagged / critical / high / medium counts per model run, for the “Predicted Drop-off Over Time” chart. Counts are score-derived from run_history so they are consistent across runs.

Query parameters

window

string

default:"all"

this_week, this_month, last_month, this_quarter, last_quarter, this_year, last_year, or all. Windows are relative to the latest run.

faculty

string

Filter by faculty.

programme

string

Filter by programme.

year

string

Filter by programme level.

cohort

string

Filter by cohort.

Response

{
  "success": true,
  "data": {
    "window": "this_month",
    "since": "2026-04-29T07:59:13.935658Z",
    "until": "2026-05-29T07:59:13.935658Z",
    "points": [
      { "run_ts": "2026-05-01T07:59:13.935658Z", "flagged": 7972, "critical": 3108, "high": 1100, "medium": 3764, "total": 10000 },
      { "run_ts": "2026-05-29T07:59:13.935658Z", "flagged": 3000, "critical": 284, "high": 716, "medium": 2000, "total": 10000 }
    ],
    "filters": {}
  },
  "message": null
}

Trend counts are score-derived (cutoffs flagged ≥0.15, high ≥0.40, critical ≥0.50), so they reconcile with the latest aggregate cards at the most recent run but use scores — not labels — for earlier runs (which have no stored label).

List At-Risk Students

GET /v1/console/student-dropoff/students

Paginated list of at-risk students for the dashboard, using the same filters as the summary cards. Defaults to the flagged population (High + Medium); narrow with risk_level.

Query parameters

faculty

string

Filter by faculty.

programme

string

Filter by programme.

year

string

Filter by programme level (e.g. 300L).

cohort

string

Filter by cohort.

risk_level

string

critical, high, medium, or low. Omit to return all flagged students (High + Medium).

period

datetime

Run timestamp. Defaults to the latest run. A historical run is re-scored from run_history (band derived from score; risk_level label is omitted for past runs).

string

Match on name, email, student id, or programme (case-insensitive).

sort_by

string

default:"drop_off_score"

drop_off_score, days_since_last_activity, or fullname.

sort_order

string

default:"desc"

asc or desc.

skip

integer

default:"0"

Records to skip.

limit

integer

default:"20"

Max records (1-100).

Example request

curl "https://staging-be.mind.miva.university/v1/console/student-dropoff/students?risk_level=critical&faculty=Computing&limit=20" \
  -H "Authorization: Bearer <access_token>"

Response

{
  "success": true,
  "data": [
    {
      "student_id": "STU001027",
      "fullname": "Jane Doe",
      "student_email": "jane.doe@miva.university",
      "faculty": "Management",
      "programme": "BSc Accounting",
      "program_level": "300L",
      "current_cohort": "202402_S1",
      "risk_level": "High",
      "risk_band": "critical",
      "drop_off_score": 0.5499,
      "score_trend_slope": 0.012,
      "trajectory": "worsening",
      "risk_reasons": ["No content completion in 14d", "Missed deadlines in last 30d"],
      "days_since_last_activity": 14,
      "missed_deadline": true,
      "n_courses": 4,
      "n_high_courses": 2,
      "latest_scored_at": "2026-05-29T07:59:13.935658Z"
    }
  ],
  "total": 284,
  "page": 1,
  "page_size": 20,
  "total_pages": 15,
  "message": null
}

risk_reasons powers the “Key signals” chips on each row.

Export At-Risk Students

GET /v1/console/student-dropoff/students/export

Download the current filtered list as CSV or XLSX. Uses the same filters, risk_level, search, and sort as the list endpoint.

Query parameters

format

string

default:"csv"

csv or xlsx.

Plus faculty, programme, year, cohort, risk_level, search, sort_by, sort_order.

Response

A file download. Columns: Student Name, Student ID, Email, Programme, Faculty, Year, Risk Band, Risk Score, Days Inactive, Missed Deadline, Key Signals, Last Scored. Filename: at_risk_students_{YYYY-MM-DD}.{csv|xlsx}. Hard cap of 50,000 rows.

Get Student Risk Profile & Trend

GET /v1/console/student-dropoff/students/{student_id}

The full risk profile for one student: current band and drivers, confidence (drop_off_score), risk trajectory, behavioural / academic / payment signals, per-course risk, and the full 24-run score trend.

Path parameters

student_id

string

required

The student’s ID (BigQuery student_id).

Example request

curl "https://staging-be.mind.miva.university/v1/console/student-dropoff/students/STU-000123" \
  -H "Authorization: Bearer <access_token>"

Response

{
  "success": true,
  "data": {
    "student_id": "STU-000123",
    "fullname": "Jane Doe",
    "student_email": "jane.doe@miva.university",
    "program": "BSc Computer Science",
    "program_level": "300L",
    "faculty": "Computing",
    "current_cohort": "202402_S1",
    "week_of_session": 9,
    "risk_level": "High",
    "risk_band": "critical",
    "drop_off_score": 0.5499,
    "confidence": 0.5499,
    "threshold": 0.7,
    "risk_reasons": ["No content completion in 14d", "Low engagement (logins below 7d band)"],
    "score_trend_slope": -0.0007,
    "trajectory": "stable",
    "risk_score_delta": 0.1083,
    "missed_deadline": true,
    "days_since_last_activity": 12,
    "engagement_level": "low",
    "missed_assessments": 3,
    "behavior": {
      "login_days_7d": 1, "login_days_30d": 4, "active_days_30d": 3,
      "learning_mins_30d": 45, "content_completed_30d": 0, "submissions_30d": 0
    },
    "academic": {
      "avg_total_score": 41.2, "avg_exam_score": 38.0, "pct_exams_completed": 0.5,
      "avg_grade_point": 1.8, "n_exams_total": 4, "best_exam_score": 55.0
    },
    "payment": {
      "payment_risk": 0.6, "payment_status": "Outstanding",
      "expected_amount": 450000.0, "amount_due": 200000.0, "balance": 200000.0
    },
    "courses": [
      {
        "course_code": "CSC401",
        "course_name": "Operating Systems",
        "risk_level": "High",
        "model_risk_level": "High",
        "drop_off_score": 0.5499,
        "review_priority": "P1",
        "factor_chips": ["↑ missed_deadline_30d", "↓ active_days_30d", "↓ login_days_30d"],
        "missed_deadline": true,
        "days_since_last_activity": 12
      }
    ],
    "trend": [
      { "model_run_id": "20260407T075913Z", "run_ts": "2026-04-07T07:59:13.935658Z", "drop_off_score": 0.4776, "risk_band": "high", "delta_vs_previous": null },
      { "model_run_id": "20260529T075913Z", "run_ts": "2026-05-29T07:59:13.935658Z", "drop_off_score": 0.5499, "risk_band": "critical", "delta_vs_previous": 0.1083 }
    ],
    "model_run_id": "20260529T075913Z",
    "model_version": "1.3.0-synth",
    "score_date": "2026-05-29",
    "latest_scored_at": "2026-05-29T07:59:13.935658Z"
  },
  "message": null
}

Notes

confidence is the model’s drop_off_score (probability of drop-off).
trajectory is derived from score_trend_slope: worsening (rising risk), improving, or stable.
risk_score_delta is the change in drop_off_score from the previous run to the latest.
engagement_level (low/medium/high) is derived from 30-day active/login days.
missed_assessments counts the student’s courses with a missed deadline.
trend contains only real runs — the warehouse pads run_history to a fixed length with empty entries, which are dropped.
404 NOT_FOUND if the student is not in the drop-off dataset.

Get Prediction History

GET /v1/console/student-dropoff/students/{student_id}/history

Chronological history of ML prediction runs for a student — risk trajectory over time, with the band and score delta at each run.

Path parameters

student_id

string

required

The student’s ID.

Response

{
  "success": true,
  "data": {
    "student_id": "STU-000123",
    "fullname": "Jane Doe",
    "score_trend_slope": -0.0007,
    "trajectory": "stable",
    "runs": [
      { "model_run_id": "20260407T075913Z", "run_ts": "2026-04-07T07:59:13.935658Z", "drop_off_score": 0.4776, "risk_band": "high", "delta_vs_previous": null },
      { "model_run_id": "20260411T075913Z", "run_ts": "2026-04-11T07:59:13.935658Z", "drop_off_score": 0.4900, "risk_band": "high", "delta_vs_previous": 0.0124 }
    ]
  },
  "message": null
}

Field reference

Field	Description
`runs[].risk_band`	Score-derived band for that run (critical/high/medium/low).
`runs[].delta_vs_previous`	Change in `drop_off_score` from the prior run; `null` for the first run.
`trajectory`	Overall direction from `score_trend_slope`.

​Get Aggregate Dashboard Stats

​Query parameters

​Example request

​Response

​Field reference

​Get Filter Options

​Response

​Predicted Drop-off Over Time

​Query parameters

​Response

​List At-Risk Students

​Query parameters

​Example request

​Response

​Export At-Risk Students

​Query parameters

​Response

​Get Student Risk Profile & Trend

​Path parameters

​Example request

​Response

​Notes

​Get Prediction History

​Path parameters

​Response

​Field reference

Get Aggregate Dashboard Stats

Query parameters

Example request

Response

Field reference

Get Filter Options

Response

Predicted Drop-off Over Time

Query parameters

Response

List At-Risk Students

Query parameters

Example request

Response

Export At-Risk Students

Query parameters

Response

Get Student Risk Profile & Trend

Path parameters

Example request

Response

Notes

Get Prediction History

Path parameters

Response

Field reference