Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/JoseOlivares19/Proyecto-PC3-JavaScript-Avanzado/llms.txt

Use this file to discover all available pages before exploring further.

The /predict/fraud-shield endpoint analyses a single financial transaction against seven behavioral and contextual risk signals and returns one of three classification labels. The underlying RandomForest model weighs transaction amount, time-of-day, prior failed attempts, customer tenure, device novelty, country risk, and recent purchase frequency to produce a fraud risk score. The response includes confidence, a full probability ranking, and a concrete action set for your fraud operations team or automated decision engine.

​
Endpoint

POST http://localhost:8001/predict/fraud-shield
Content-Type: application/json

​
Request Body

​
monto
float
required
Transaction amount in local currency. Must be between 1 and 20000.
​
hora_24
integer
required
Hour of the day the transaction was initiated, in 24-hour format. Must be between 0 and 23.
​
intentos_previos
integer
required
Number of previous failed transaction attempts by this customer in the current session. Must be between 0 and 10.
​
antiguedad_cliente_meses
integer
required
Customer account age in months. Must be between 0 and 120. Newer accounts carry higher risk weight.
​
dispositivo_nuevo
integer
required
Whether the transaction originates from a device not previously seen for this account:
  • 0 — Known device
  • 1 — New / unrecognised device
​
pais_riesgo
integer
required
Risk classification of the transaction’s originating country:
  • 0 — Low-risk country
  • 1 — High-risk country
​
compras_ultima_hora
integer
required
Number of purchase attempts made by this account in the last 60 minutes. Must be between 0 and 20. Values above 5 contribute additional risk.

​
Example Request

curl -X POST http://localhost:8001/predict/fraud-shield \
  -H "Content-Type: application/json" \
  -d '{
    "monto": 8750.00,
    "hora_24": 2,
    "intentos_previos": 3,
    "antiguedad_cliente_meses": 1,
    "dispositivo_nuevo": 1,
    "pais_riesgo": 1,
    "compras_ultima_hora": 7
  }'

​
Example Response

{
  "caso": "FraudShield",
  "prediccion": "FRAUDE_PROBABLE",
  "confianza": 0.9467,
  "ranking": [
    { "clase": "FRAUDE_PROBABLE",    "probabilidad": 0.9467 },
    { "clase": "REVISION_MANUAL",    "probabilidad": 0.0421 },
    { "clase": "TRANSACCION_SEGURA", "probabilidad": 0.0112 }
  ],
  "recomendaciones": [
    "Bloquear temporalmente la operación.",
    "Solicitar autenticación reforzada.",
    "Registrar alerta para revisión del analista."
  ],
  "entrada": {
    "monto": 8750.0,
    "hora_24": 2,
    "intentos_previos": 3,
    "antiguedad_cliente_meses": 1,
    "dispositivo_nuevo": 1,
    "pais_riesgo": 1,
    "compras_ultima_hora": 7
  }
}

​
Response Fields

​
caso
string
Human-readable model identifier. Always "FraudShield" for this endpoint.
​
prediccion
string
The top predicted fraud classification. One of:
LabelMeaning
FRAUDE_PROBABLEHigh fraud risk — block and escalate immediately
REVISION_MANUALModerate risk — request additional customer verification
TRANSACCION_SEGURALow risk — authorise and continue passive monitoring
​
confianza
float
Probability assigned to the top predicted class, in the range 0.0 to 1.0.
​
ranking
array of objects
Probability distribution across all three fraud labels, sorted in descending order of probability.
​
recomendaciones
array of strings
Action set appropriate to the predicted label. For FRAUDE_PROBABLE this includes blocking and escalation steps; for REVISION_MANUAL it includes verification steps; for TRANSACCION_SEGURA it authorises the operation.
​
entrada
object
Echo of the validated request payload as parsed by Pydantic. Use this for audit logging without re-serialising the original request.
dispositivo_nuevo and pais_riesgo are binary integer flags (0 or 1). Boolean true/false values are not accepted and will produce a 422 Unprocessable Entity error.

Build docs developers (and LLMs) love

Get started for freeTalk to us