House Price Prediction – Analytic Approach
Analytic Approach from bussiness understanding to Analysis design software.
I. Analytic Approach (Data-Oriented Thinking)
1. Purpose of the Analytic Approach
Once the business problem is clearly understood, the next step is to determine how data analysis can answer the question.
The analytic approach involves:
- Clarifying the type of question being asked
- Selecting the most appropriate analytical method
2. Mapping the Question to an Analytic Approach
The refined analytical question is:
- “Given the characteristics of a house, what price should we estimate?”
This question requires:
- A numerical output
- Prediction based on historical patterns
This is a Regression Problem
3. Selecting the Type of Analysis
| Question Type | Analytic Approach |
|---|---|
| Predict a numeric value | Regression |
| Yes / No decision | Classification |
| Discover patterns | Clustering |
| Describe trends | Descriptive Analysis |
For house price prediction:
- Regression analysis is the most appropriate
- Machine Learning may be used to capture complex relationships
4. Why Machine Learning May Be Used
Machine Learning can:
- Identify non-linear relationships between features and price
- Improve prediction accuracy compared to simple rules
However:
- ML is a tool, not the objective
- The choice of model comes after this step
II. Comparison of the Two Approaches
| Problem Approach | Analytic Approach |
|---|---|
| Focuses on goals and value | Focuses on methods |
| Business-driven | Data-driven |
| Defines what success means | Defines how to measure it |
| No models mentioned | Model types identified |
III. Analysis Design
1. Type of Data Science Problem
| Business Need | Analytics Interpretation |
|---|---|
| Estimate a numerical house price | Predict a continuous numeric value |
This is Supervised Learning – Regression
2. Frome Input to Output Mapping
| Role | Business Meaning | Data Representation |
|---|---|---|
| Input (X) | Property attributes | Area, location, rooms, year built, etc. |
| Output (y) | House value | Sale price |
Mathematically:
\[Price=f(Property,Location,Market)\]3. What Is the Model Expected to Do?
The model must:
- Learn the relationship between house characteristics and historical prices
- Generalize to unseen houses
- Produce a numeric price estimate
Not:
- Classify houses
- Rank houses
- Recommend houses
4. Success Metrics (Analytics View)
Business Success Criteria: “Average prediction error below an acceptable threshold”
Mapping: | Business Concern | Analytics Metric | | —————————- | —————————— | | How far off is the estimate? | MAE (Mean Absolute Error) | | Penalize large mistakes | RMSE (Root Mean Squared Error) | | Relative performance | R² score |
- Primary metric: MAE
- Secondary: RMSE
5. Baseline Definition
Business wants to know: “Is this better than current practice?”
Baseline:
- Mean house price
- Or price per m² × area
| Model | Purpose |
|---|---|
| Simple heuristic | Business benchmark |
| ML model | Value-added comparison |
6. Output Form
| Field | Description |
|---|---|
| predicted_price | Estimated house value |
| confidence_interval (optional) | Uncertainty |
7. Deployment Implication (from Analytics)
- User inputs numeric & categorical features
- System returns a number
- UI must support structured input
8. Analytic Design Summary
This project is a supervised regression problem where historical house transaction data will be used to train a model that predicts continuous house prices from property, location, and market features. The solution will be evaluated primarily using MAE and RMSE against a baseline pricing heuristic.