ForgeAI Insights UI

Documentation for the Insights Dashboard

Overview:

The insights dashboard is designed as a dynamic tool that adapts based on the specific machine learning model used for regression or classification. It visualizes how various input metrics influence predictions, allowing for a real-time, adaptable view of business-critical insights. The type of data displayed can change depending on the model set up to either classify or apply regression to key metrics.

Key Sections:

1. Model Selector:
   • The left-hand menu allows users to toggle between different existing models.
   • Depending on the machine learning model and the use case, this menu may reflect different drivers relevant to the business (e.g., customer churn, equipment status, etc.).
2. Feature Data Inputs:
   • The panel shows the range of feature data contributing to the current analysis, from a minimum to a maximum value, which varies based on the model in use.
   • Depending on whether the model is performing regression or classification, the range could represent various metrics, such as number of users during peak times, or specific user behaviors being tracked.
3. Identifier or Metric value inputs:
   • Displays an identifier or metric value related to the current configuration.
   • This metric is model-dependent and could represent an aggregate of various inputs or a calculated score that influences predictions (e.g., user engagement index, sensor data, or machine activity scores).
   • As the machine learning model changes, this dashboard metric could reflect anything from system health to customer segmentation scores.
4. Predictions Chart:
   • The chart labeled "Predictions" visualizes the relationship between the input metrics (e.g., active users) and the predicted outcomes.
   • Y-axis: This represents the target outcome, which may vary depending on the model. For example:
     • In a classification model: The Y-axis might indicate probabilities for different classes (e.g., high/low revenue potential, equipment status, user churn risk).
     • In a regression model: It could represent continuous predicted values such as revenue, machine uptime, or customer satisfaction scores.
   • X-axis: Represents the input variables (such as active users, machine uptime, or production data), with the range and units adapting to the dataset being analyzed.
   • The visual spikes or trends on the chart highlight the prediction points where certain input levels trigger higher probabilities or predicted outcomes, such as revenue growth, equipment failure, or increased user engagement.

Model-Specific Insights:

• Classification Models:
  • The chart can highlight when certain features (e.g., number of active users) contribute to a classification outcome (such as high vs. low engagement, churn risk levels, or operational status).
  • The spikes represent thresholds where the model has high confidence in predicting a specific class, such as identifying users most likely to churn when active user counts cross a certain threshold.
• Regression Models:
  • The chart may predict continuous outcomes like revenue, energy consumption, or production output based on the input data.
  • The trends reflect how different input variables (such as user activity or sensor readings) impact a target value over time (e.g., revenue predictions or machine performance metrics).

Usage:

This dashboard is designed to provide flexibility depending on the machine learning model applied. Whether the goal is classification or regression, it empowers teams to visualize key trends and predictions in real-time, enabling data-driven decision-making.

Potential Actions:

• Model-Specific Adjustments: Depending on the machine learning model in use, teams can focus on optimizing different aspects of their business. For example:
  • For classification models (e.g., churn prediction): Focus on users at risk of churn and implement retention strategies.
  • For regression models (e.g., revenue predictions): Align marketing and sales strategies with projected revenue trends.
• Customizable Predictions: The dashboard dynamically updates predictions, helping businesses fine-tune operations by acting on insights such as predictive maintenance, customer engagement, or revenue predictions.

Conclusion:

This insight dashboard is a versatile tool capable of adapting to different machine learning models and use cases. By shifting between classification and regression models, users gain the flexibility to address diverse business challenges, from operational efficiency to revenue growth or customer retention.