Akkio Docs
  • Akkio Documentation
    • Akkio FAQ
  • Account and Settings
    • Team Settings
    • Organization Settings
    • Account Settings
    • Role Based Access Control
  • Demo Models
    • Demo Models
      • Lead Scoring
      • Retail Sales Forecasting
      • Predict Credit Card Fraud
      • Identify Customer Churn
  • Setting up Integrations
    • Connecting Data
    • Airtable (Beta)
    • Google Ads (Beta)
    • Google Analytics 4 (Beta)
    • Google BigQuery
    • Google BigQuery (Service Account)
    • Google Sheets
    • HubSpot (Beta)
    • MariaDB (Beta)
    • MongoDB (Beta)
    • MySQL (Beta)
    • PostgreSQL (Beta)
    • Redshift (Beta)
    • Salesforce
    • Akkio Data Chat for Slack
    • Snowflake (Username / Password) (Beta)
    • Zapier
  • Prepare your Data
    • Prepare
      • Chat Data Prep
      • Clean
      • Merge & Fuzzy Merge
      • Table View
      • Pivot View
      • Deploying Chat Data Prep
  • Explore
    • Chat Explore
    • Chart Types
  • Building a Model
    • Predict
      • Insights Report - Classification
      • Insights Report - Regression
    • Forecasting
      • Insights Report - Forecasting
    • Model Types
  • Deploying a Model
    • Deploy
      • Google BigQuery
      • Google Sheets
      • HubSpot (Beta)
      • PostgreSQL (Beta)
      • Salesforce
      • Snowflake (Beta)
      • Web App
      • Zapier
  • REPORTING AND SHARING
    • Reports
    • Dashboards
  • REST API
    • API Introduction
      • Quickstart
    • API Options
      • cURL Commands
      • Python Library
      • Node.js Library
    • API FAQ
  • Rest API (v2)
    • Documentation
Powered by GitBook
On this page
  • Model Selection
  • Training Options - Predict
  • Predict Fields
  • Training Mode
  • Advanced

Was this helpful?

  1. Building a Model

Predict

PreviousChart TypesNextInsights Report - Classification

Last updated 1 year ago

Was this helpful?

Predict is the core of the Akkio platform. Providing AI-powered forecasting for your BI needs. Here, you build an AI model to make predictions on your connected data.

Model Selection

This will take you to a selection page with the following options:

  • Predict - Predict outcomes for data based on categories and numerical values. These predictions can be boolean or numerical. Examples are lead scoring, housing prices, employee attrition, etc.

Training Options - Predict

Predict Fields

Select what field you want to predict; each flow can only predict one column at a time. Select the 'Ignore' tab to ignore any fields that may lead to erroneous models. These can be anything from calculated to status columns.

Training Mode

The “Training Mode” section gives you several options for model training:

  • Fastest

  • High quality

  • Higher quality

  • Production

Advanced

  • Prediction Threshold - Limits the amount of returned predictions to a specified confidence. Defaults to 0, so all predictions are returned.

  • Automatic Retraining - Retrains the model when the input dataset changes. This can ensure that when live data is connected, the model is always retrained on the most current version of the dataset.

Once happy with your settings, you can press 'Create Predictive Model.'

- Forecast time-dependent outcomes based on historical data with time information. See sales projections based on time of year, temperature ranges based on season, energy usage, etc.

The quality options are ordered in the time taken to train. This won’t necessarily lead to higher accuracy due to a phenomenon known as “,” when a machine learning model only learns the training data well but does not generalize to new data.

Forecast (Time Series)
overfitting