Survival analysis (time-to-event)

Predict how long until an event happens, allowing for cases where it has not happened yet (censoring). The dataset is a single CSV where every row is one subject: covariate columns, a duration column and an event indicator column (1 = event observed, 0 = censored).

Folder layout

/data/shared/survival/
└── survival.csv

Data CSV

age,anaemia,creatinine_phosphokinase,diabetes,ejection_fraction,high_blood_pressure,platelets,serum_creatinine,serum_sodium,sex,smoking,time,DEATH_EVENT
48,0,315,0,40,1,275841,1.6,131,0,1,225,0
78,0,3964,1,62,0,315081,0.9,135,0,0,5,1
64,0,2656,0,45,0,372021,3.0,142,0,1,13,1

Column	Meaning
covariate columns	Declared in `schema` with their SQL type.
duration column (`time` here)	Name it `time`. Training reads only a column named `time`: a different name set with `time_column` ingests, but training cannot find it. Every value must be a number greater than 0. The ingestor accepts `0` and only warns on empty values, but training rejects both. Declared in `schema` like any other column.
event indicator (`DEATH_EVENT` here)	The label column, named by `label.column` and not declared in `schema`. Must be `0` (censored) or `1` (event). Other codes ingest, but training rejects them.

Ingest with the CLI

Pass the CSV file, or a folder that holds only that CSV. Ingest the training and the test split separately:

tracebloc data ingest ./survival_train.csv --task time_to_event_prediction --name heart_failure_survival_train --intent train \
  --label-column DEATH_EVENT --time-column time --label-policy bucket
tracebloc data ingest ./survival_test.csv --task time_to_event_prediction --name heart_failure_survival_test --intent test \
  --label-column DEATH_EVENT --time-column time --label-policy bucket

--time-column time and --label-policy bucket are the defaults. The CLI infers a SQL type for every column, so --schema is optional. To set the types yourself, pass --schema 'col:TYPE,col:TYPE' and quote it, because a type such as VARCHAR(64) contains shell characters. --schema replaces the inferred schema: list every feature column you want stored, not just the ones you retype. Including the label column in --schema is fine: the ingestor drops it from the stored schema. The CLI has no flags for time_unit and event_indicator: they are optional and can only be set in ingest.yaml.

ingest.yaml

apiVersion: tracebloc.io/v1
kind: IngestConfig
category: time_to_event_prediction
table: heart_failure_survival_train
intent: train
csv: /data/shared/survival/survival.csv
time_column: time          # the default; training reads only a column named time
time_unit: days
event_indicator:
  event: 1
  censored: 0
schema:
  age: INT
  anaemia: INT
  creatinine_phosphokinase: INT
  diabetes: INT
  ejection_fraction: INT
  high_blood_pressure: INT
  platelets: FLOAT
  serum_creatinine: FLOAT
  serum_sodium: INT
  sex: INT
  smoking: INT
  time: INT                # the duration column is declared like any other
label:
  column: DEATH_EVENT
  policy: bucket

Field	Required	Meaning
`csv`	yes	Path to the data CSV.
`schema`	yes (with the CLI, inferred for you)	Covariates plus the duration column, with SQL types. Every schema column must exist in the CSV.
`label`	yes, object form	`column` names the event indicator; `policy` is required. `bucket` makes the platform receive hash buckets, not raw values — stored rows keep the raw indicator that training reads. `passthrough` sends raw values; use it only with a compliance sign-off. With `bucket`, the experiment’s class counts show hash bucket ids instead of event and censored counts.
`time_column`	no	Name of the duration column. Default `time`. Matched exactly as written. Keep the default: training reads only `time`.
`time_unit`	no	`days`, `weeks`, `months` or `years`. Recorded so combined datasets are not compared across units.
`event_indicator`	no	`{event: 1, censored: 0}`. Recorded so a flipped convention across combined datasets is caught. It does not change the codes training accepts, which are always `1` and `0`.
`columns`	no	Per-column `unit` and `ordinal` facts.
`data_id`, `spec.csv_options`	no	See the contract.

What the ingestor checks

Check	Rejects
Time to event	A CSV without the duration column (exact name); non-numeric duration values; negative durations. Empty durations produce a warning.
Data types	Duplicate headers, schema columns missing from the CSV, values that do not match their declared type, unknown SQL types.

Plus the checks every ingest runs. There is no label-diversity check.

Sample dataset

The template ships 30 synthetic rows shaped like a heart-failure clinical dataset (no real patients): eleven covariates (age, anaemia, creatinine_phosphokinase, diabetes, ejection_fraction, high_blood_pressure, platelets, serum_creatinine, serum_sodium, sex, smoking), the duration time in days and the event indicator DEATH_EVENT. The covariates are linearly independent on purpose — a Cox proportional-hazards fit needs that. The rows shown above are its first rows, and the ingest.yaml above ingests it as is; only csv: changes to wherever you staged the file.

Model settings for this dataset

When you link a model, the platform checks num_feature_points against the dataset (see model parameters by task). It must equal the number of stored columns minus the event indicator, so the duration column counts. For the sample, that is 12, which matches the model-zoo default. output_classes is not checked for this task. Feature columns reach the model sorted by column name, whatever order the CSV or schema uses. The lifelines and scikit-survival models, such as cox_ph, need the matching SDK extra. Without it, the local model check fails with No module named 'lifelines':

pip install "tracebloc[lifelines,scikit-survival]>=1.2.47"

Next steps

Ingest the data: Prepare Data
Shared rules for every template: Dataset templates

Overview

Create a Use Case

Join a Use Case

Environment Setup

Tools & Help

Releases

Survival analysis (time-to-event)

Folder layout

Data CSV

Ingest with the CLI

ingest.yaml

What the ingestor checks

Sample dataset

Model settings for this dataset

Next steps

​Folder layout

​Data CSV

​Ingest with the CLI

​ingest.yaml

​What the ingestor checks

​Sample dataset

​Model settings for this dataset

​Next steps

Folder layout

Data CSV

Ingest with the CLI

ingest.yaml

What the ingestor checks

Sample dataset

Model settings for this dataset

Next steps