A company wants to predict house sale prices using multivariable linear regression. The dataset includes features such as lot size, living area, non-living area, bedroom count, bathroom count, year built, and postal code. Which procedure should a machine learning specialist use to eliminate irrelevant features and reduce model complexity?

To eliminate irrelevant features and reduce model complexity in a regression task, it's crucial to identify features that have a strong relationship with the target variable (house sale price). Computing the correlation of each feature with the target sale price directly measures this relationship. Features with low correlation contribute little to predicting the target and can be removed without significantly impacting model performance, thereby reducing complexity. Plotting histograms and computing standard deviations primarily helps understand feature distributions and spread, but doesn't directly indicate relevance to the target. Removing features based solely on high or low variance might discard important predictors or retain irrelevant ones. Creating a heatmap of inter-feature correlations helps identify multicollinearity among independent variables, which is useful for model stability, but doesn't directly assess a feature's relevance to the target variable.