Basic Predictive Analysis

Using Patterns to Forecast Outcomes

Predictive analysis uses historical patterns to forecast future outcomes. It's more complex than descriptive analysis but increasingly accessible with AI.

What Prediction Means

Not Crystal Balls

Predictions are probabilistic, not certain. They're educated guesses based on patterns.

Based on Assumptions

Predictions assume the future resembles the past. When conditions change, predictions break.

Useful Despite Limitations

Imperfect predictions are often better than no predictions. They provide a baseline for planning.

Types of Prediction

Forecasting

Predicting future values of time-series data.

Examples:

  • Next month's sales
  • Website traffic tomorrow
  • Demand next quarter

Methods:

  • Trend extrapolation
  • Seasonal models
  • Moving averages

Classification

Predicting which category something belongs to.

Examples:

  • Will this customer churn? (Yes/No)
  • Is this email spam?
  • What segment does this user belong to?

Regression

Predicting a continuous value.

Examples:

  • What will this customer spend?
  • How long will this process take?
  • What price should we set?

Simple Forecasting Methods

Naive Methods

Last value: Tomorrow = today Average: Future = historical average Seasonal naive: Same period last year

Simple but provide baselines to beat.

Trend Extrapolation

Extend the current trend into the future.

Assumption: Growth/decline continues at same rate.

Limitation: Trends don't continue forever.

Moving Averages

Average of recent periods. Smooths out noise.

Simple moving average: Average of last N periods Weighted moving average: Recent periods weighted more

Seasonal Adjustment

Account for predictable seasonal patterns.

Example: Retail spikes in December. Adjust forecasts to expect this.

What Drives Predictions

Historical Patterns

Past behavior is the primary input for prediction.

Leading Indicators

Variables that predict outcomes:

  • Website visits predict sales
  • Economic indicators predict industry performance
  • Customer behavior predicts churn

Features (Predictors)

In more complex models, multiple variables combine to predict outcomes.

Evaluating Predictions

Accuracy Metrics

Mean Absolute Error (MAE): Average absolute difference between predicted and actual.

Root Mean Square Error (RMSE): Penalizes large errors more.

Percentage errors: MAPE, etc.

Prediction vs. Actual

Compare predictions to what actually happened. Track over time.

Prediction Intervals

Not just point estimates — how confident are we? Ranges are more honest than single numbers.

Limitations of Prediction

Data Limitations

Predictions are only as good as input data. Missing data, errors, and biases carry through.

Changing Conditions

Models trained on historical data assume conditions continue. Market shifts, competition, disruptions break predictions.

Complexity

Real-world outcomes have many causes. Simple models capture some, not all.

Overfitting

Models can fit historical data perfectly but fail on new data. Simpler models often generalize better.

When to Use Simple vs. Complex Methods

Simple Methods Work When

  • Patterns are clear and stable
  • Data is limited
  • Quick estimates are needed
  • Transparency matters

Complex Methods Help When

  • Large datasets available
  • Multiple factors interact
  • Patterns are subtle
  • Accuracy is critical

AI for Prediction

What AI Does Well

  • Pattern recognition in complex data
  • Handling many variables
  • Automating forecasting
  • Explaining prediction drivers

How to Use It

Ask AI to:

  • Identify patterns that might predict outcomes
  • Create simple forecasts
  • Explain what drives predictions
  • Evaluate prediction approaches

AI Prompt: Simple Forecast

Help me create a forecast.

Historical data: [Paste or describe time series]
What I'm forecasting: [The variable]
Forecast horizon: [How far ahead]
Any known factors: [Seasonality, events, etc.]

Please:
1. Identify patterns in the historical data
2. Create a simple forecast
3. Explain the methodology
4. Estimate confidence/uncertainty
5. Note assumptions and limitations

AI Prompt: Prediction Drivers

Help me understand what predicts this outcome.

Outcome I want to predict: [What you're predicting]
Available data: [Variables you have]
Sample: [Paste data if available]

Please:
1. Identify likely predictor variables
2. Explain the expected relationships
3. Suggest how to test these relationships
4. Note what data might be missing
5. Recommend next steps

What's Next

Let's get specific about using AI for analysis.

Next chapter: Using AI for analysis — practical techniques for every task.

Finding RelationshipsUsing AI for Analysis