Skip to content

Latest commit

 

History

History
49 lines (38 loc) · 4.25 KB

README.md

File metadata and controls

49 lines (38 loc) · 4.25 KB
Raw

Housing Project

Develop predictive models to estimate housing prices based on various influencing factors. I implemented this project in both R and in Python.

Portfolio Project: House Prices

Objective: The goal of my analysis is to predict the price of homes based on various features of housing such as number of bedrooms, house area, furnishing, etc.

Data: The Housing Prices dataset from Kaggle contains 545 observations with 13 features. These features include the median value of homes (target variable) and attributes such as average number of bedrooms, bathrooms, area square footage, etc.

Skills

  • R programming
  • Python
  • Linear Regression
  • Predictive and statistical analysis
  • Visualizations

Exploratory Analysis (EDA):

  1. Data Cleaning:
    I checked the data for missing values and duplicates. There were no nulls or duplicates in this dataset. I checked the data types and found that some of the predicting variables were categorical variables. I converted the categorical data using as.factor() to numerical.
  2. Relationships: My EDA revealed that features such as the number of bathrooms and area, had moderate correlations with housing prices. The remaining correlations seem to be moderate to weak. Visualizations included histograms, box plots, and scatter plots to understand the distributions and relationships between variables. I evaluated the boxplot of each categorical variable against the dependent variable price that showed the medians. Created scatterplots to observe the linear relationships between price and each predicting variable. I took note of the trends and direction for each. I created a correlation matrix to check the correlation between the variables. Most of the correlations were weak, with only a few moderately strong correlations between price and the independent variable. I did not observe any high correlations between the predicting variables, suggesting against multicollinearity.

Analysis:

  1. Model Fitting : Split the data into a training set and testing set. I chose to fit the data using Linear regression due to its simplicity and interpretability. The full model included all available features, while the reduced model included only the most significant predictors at alpha 0.01. Model Evaluation: I compared the two models using r^2 and the adjusted r^2 value to account for the additional predictors. I computed the mae, rsme and mse value of the two models for model evaluation and comparison.

Findings: The P-value of the f-statistic is 2.2e-16, which is less than the alpha level so we can reject the null. Concluding that, the additional predictors in the full model significantly improves the model. R^2 is the proportion of variance explained by the predictors. Adj r^2 considers the additional variables. The full model has a higher adj r^2 value than the reduced model. It explains 68.3% of the variability in price. That's about 7% more than the reduced model.

  1. Outliers & Multicollinearity : I checked for significant outliers using cook's distance with a threshold of 1, and found no significant outliers. I checked for multicollinearity among the predictors using VIF and found that no values exceeded the threshold, therefore no evidence of multicollinearity.

  2. Prediction Obtained very large rsme and mse values for the prediction model. Overall, the full model that includes all of the predictors was able to predict 67% of the variability in the data. This indicates that the additional predictors in the full model contribute significantly to explaining the variation in housing prices. It is also a better model than the reduced model, by about 7% since it explains more of the variability in price.

  3. Conclusion The full model, with its higher adjusted R-squared value, demonstrates a better fit to the data compared to the reduced model. Specifically, the full model explains 68.3% of the variability in housing prices. While the full model provides a higher adjusted R^2 value than the reduced model, the high RMSE and MSE values suggest that there is room for improvement. The high RMSE and MSE values suggests possible issues like overfitting, outliers, multicollinearity, or too many predictors (overly complex model). Further refinements are necessary to develop a more reliable and robust predictive model for housing prices.