Post

House Price Prediction – Analytic Approach

Analytic Approach from bussiness understanding to Analysis design software.

House Price Prediction – Analytic Approach

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 TypeAnalytic Approach
Predict a numeric valueRegression
Yes / No decisionClassification
Discover patternsClustering
Describe trendsDescriptive 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 ApproachAnalytic Approach
Focuses on goals and valueFocuses on methods
Business-drivenData-driven
Defines what success meansDefines how to measure it
No models mentionedModel types identified

III. Analysis Design

1. Type of Data Science Problem

Business NeedAnalytics Interpretation
Estimate a numerical house pricePredict a continuous numeric value

This is Supervised Learning – Regression

2. Frome Input to Output Mapping

RoleBusiness MeaningData Representation
Input (X)Property attributesArea, location, rooms, year built, etc.
Output (y)House valueSale 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
ModelPurpose
Simple heuristicBusiness benchmark
ML modelValue-added comparison

6. Output Form

FieldDescription
predicted_priceEstimated 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.

This post is licensed under CC BY 4.0 by the author.