Skip to content

PlatformAnalyse

Predictors

From past data to an expected value, with versions under your control

A predictor estimates what to expect from your data: a trend with its band, an anomaly, a value you do not record, or the probability that an event starts. You build it from a query, in four steps.

Book a demoThe whole platform
Predictor · Sales by areaExample
Kind
Forecast
What
Daily sales, North area
From where
Query «Sales by day»
Active version
v1
Candidate
v2, waiting

A person promotes the candidate version.

Illustrative example

Four kinds, one way of working

  • Four kinds

    Anomaly detection, forecast with a band, virtual sensor and probability. Each kind is a module of its own in the Lab; probability also needs Analisi.

  • From a query, in four steps

    What to estimate, where the data comes from, when to train and apply, and an optional alarm. Probability adds the event.

  • Versions you decide on

    Versions are not edited. The first becomes active; later ones stay candidates until a person promotes or rejects them.

  • Your own model, if you need it

    You can upload a model trained elsewhere as a file: it is tried before it becomes a version.

How it works

  1. 01

    Start from a query

    Choose the kind and, with «Cosa» and «Da dove» (what and from where), point to the series to estimate: for example the daily sales of a product category.

  2. 02

    Train and apply

    With «Quando» (when) you choose when the predictor is trained and applied; if you like, add an alarm.

  3. 03

    Promote the version

    A new version stays a candidate until you decide whether to promote it. The last ten retired ones stay within reach for a rollback.

In detail

Kinds
Anomaly detection, Forecast, Virtual sensor, Probability: separate modules in the Lab's Predictors section.
Steps
Cosa, Da dove, Quando, Allarme; probability adds Evento.
Versions
Immutable: the first becomes active, the others are candidates to promote or reject.
Your own model
Uploaded as a file and tried before it becomes a version.
Without training
A ready-made forecast and the detection of its anomalies run with no training, under the modules of those kinds.
AI on text
From a text column, such as customer notes or reports, it sorts each row among the labels you choose, summarises it or extracts the fields you name, into a table. It runs in the cloud and starts once an administrator approves it.
Athena
Can propose a predictor; you approve it in «Da approvare».

Also in Analyse

Frequently asked questions

Muvia does not state an accuracy: it depends on your data. That is why a new version stays a candidate until a person promotes it.

Bring us a question you can't answer today.

We start from a real question your business has and walk the path from source to dashboard with your systems, not a demo dataset.

Book a demo