Machine Learning Project: Predicting House Prices Syllabus

Project: Predicting House Prices syllabus — full chapter and topic list

Expand any chapter to see its topics and sub-topics. This is the whole examinable outline for Project: Predicting House Prices in Machine Learning, not a summary of it.

1. 1

   Define the Problem

   2 topics

   • Identify the Objective
     • Determine the target variable
     • Specify the prediction task
   • Understand the Business Context
     • Gather domain knowledge
     • Identify stakeholders
2. 2

   Collect Data

   2 topics

   • Identify Data Sources
     • Public datasets
     • Proprietary datasets
   • Gather Data
     • Download datasets
     • Merge datasets
3. 3

   Data Preprocessing

   3 topics

   • Data Cleaning
     • Handle missing values
     • Remove duplicates
   • Feature Engineering
     • Create new features
     • Encode categorical variables
   • Data Transformation
     • Normalize/Standardize data
     • Split data into training and testing sets
4. 4

   Model Selection

   2 topics

   • Choose Algorithms
     • Linear Regression
     • Decision Trees
     • Random Forest
     • Gradient Boosting
   • Model Training
     • Train models on training data
     • Tune hyperparameters
5. 5

   Model Evaluation

   2 topics

   • Assess Model Performance
     • Evaluate on testing data
     • Use metrics like RMSE, MAE
   • Compare Models
     • Select the best performing model
6. 6

   Model Deployment

   2 topics

   • Prepare for Deployment
     • Serialize the model
     • Develop API for model inference
   • Deploy the Model
     • Host the model on a server
     • Monitor model performance
7. 7

   Model Maintenance

   2 topics

   • Monitor Model Performance
     • Track prediction accuracy
     • Update model with new data
   • Retrain Model
     • Periodically retrain with updated data
     • Re-deploy updated model

Project: Predicting House Prices flashcards for Machine Learning

18 of 50 cards from the Project: Predicting House Prices deck — real questions with worked answers.

1. In a machine learning project, what does 'Identifying the Objective' mean, and what is the objective for the House Prices project?

   It means defining the precise problem the model must solve and the target variable to predict. For House Prices, the objective is to predict the continuous sale price of a house from its features — a supervised regression task.

2. Why is predicting house prices classified as a regression problem rather than classification?

   Because the target variable (price) is a continuous numeric value, not a discrete category. Regression predicts continuous outputs; classification predicts discrete labels.

3. What is the difference between a business objective and a machine learning objective in a project?

   The business objective is the real-world goal (e.g., help buyers/sellers price homes accurately). The ML objective is the technical translation of it (e.g., minimize prediction error $\text{RMSE}$ on house price).

4. In 'Understanding the Business Context' for house price prediction, who are typical stakeholders and how would they use the model?

   Stakeholders include real-estate agents, buyers, sellers, lenders, and appraisers. They use predicted prices to set listing prices, make offers, assess loan collateral, and detect over/under-priced listings.

5. What is a 'success metric' agreed upon in the business-context stage, and give an example for house prices.

   A quantitative target that defines acceptable model performance, e.g., 'mean absolute error under \$15,000' or 'predictions within 10% of actual price for 80% of homes'.

6. What are 'structured' vs 'unstructured' data sources, with examples for a housing dataset?

   Structured data is organized in tables/rows-columns (e.g., square footage, bedrooms, year built). Unstructured data has no fixed schema (e.g., listing photos, free-text descriptions).

7. Name three categories of data sources you might identify for a house-price model.

   Internal records (past sale transactions), public/government data (tax records, census, geographic data), and third-party/API data (Zillow-like listings, mortgage rates, neighborhood crime/school ratings).

8. What is the difference between primary and secondary data sources?

   Primary data is collected first-hand for the specific purpose (e.g., your own sales database). Secondary data is collected by others and reused (e.g., census or open real-estate datasets).

9. In the 'Gather Data' step, what is the difference between batch and streaming (real-time) data collection?

   Batch collection gathers data in large chunks at intervals (e.g., nightly export of sales). Streaming collection ingests data continuously as it is generated (e.g., live listing updates).

10. What is data 'sampling' and why might you sample when gathering a large housing dataset?

    Sampling selects a representative subset of the full data. It reduces compute and storage cost while preserving the population's statistical properties, enabling faster experimentation.

11. Define stratified sampling and why it helps in house-price data.

    Stratified sampling divides data into subgroups (strata, e.g., price ranges or neighborhoods) and samples proportionally from each, ensuring the sample mirrors the population's distribution and avoids bias toward common groups.

12. What are the main goals of the 'Data Cleaning' step?

    To handle missing values, remove duplicates, fix inconsistent formats, correct errors, and address outliers — producing a reliable, consistent dataset for modeling.

13. List three common strategies for handling missing values in a housing dataset.

    Deletion (drop rows/columns with missing data), imputation (fill with mean/median/mode or a model-based estimate), and using a 'missing' indicator flag to mark absence as informative.

14. Why is the median often preferred over the mean for imputing missing house prices or areas?

    Because price/area distributions are right-skewed with outliers; the median is robust to extreme values, whereas the mean is pulled toward them, distorting the imputed value.

15. What is an outlier, and name one statistical method to detect outliers in house data.

    An outlier is an observation far from the rest of the data. The IQR method flags values below $Q_1 - 1.5\,\text{IQR}$ or above $Q_3 + 1.5\,\text{IQR}$, where $\text{IQR} = Q_3 - Q_1$.

16. Give the z-score formula used to detect outliers and the common threshold.

    $$z = \frac{x - \mu}{\sigma}$$ Values with $|z| > 3$ are typically treated as outliers.

17. What is Feature Engineering and why is it important for house-price models?

    It is the creation of new input features from raw data to better expose patterns to the model. Good features (e.g., price-per-sqft, house age) often improve accuracy more than changing the algorithm.

18. Give two examples of engineered features for a house-price dataset.

    Age of house = current year − year built; total bathrooms = full + 0.5 × half baths; price per square foot; or distance to city center derived from latitude/longitude.

See more Project: Predicting House Prices flashcards →

Planning Project: Predicting House Prices for Machine Learning

Project: Predicting House Prices (Machine Learning) FAQ