{"media":{"disableImageZooming":false},"sections":[{"weight":1,"colorBackground":"#96ceb4","colorText":"#000000","title":"Adjusting Parameters","description":"

<\/p>\n\n

If a model is underperforming, one approach is to adjust its parameters. For example, tweaking the learning rate in machine learning models or adjusting the weight given to different variables can improve prediction accuracy.<\/p>\n\n

Example: A company using a regression model finds that certain factors (like holiday seasonality) are underweighted. By increasing the weight of those factors, the model\u2019s predictive accuracy improves.<\/p>\n"},{"weight":1,"colorBackground":"#96ceb4","colorText":"#000000","title":"Incorporating New Data","description":"

<\/p>\n\n

Predictive models should be continuously updated with new data to ensure they remain relevant. For instance, incorporating recent trends or new customer behaviors, such as changes in purchasing patterns after a global event (e.g., the COVID-19 pandemic), can significantly improve model performance.<\/p>\n\n

Example<\/strong>: A retailer finds that customer purchasing behavior shifted drastically post-pandemic. By incorporating this new data, they update their predictive models to better forecast future demand, improving stock management and customer satisfaction.<\/p>\n"},{"weight":1,"colorBackground":"#96ceb4","colorText":"#000000","title":"Changing Models","description":"

<\/p>\n\n

If a model consistently fails to provide accurate forecasts, it might be necessary to switch to a different model type. For instance, if a linear regression model doesn\u2019t capture non-linear patterns, switching to a machine learning algorithm like random forests<\/strong> or neural networks<\/strong> might help.<\/p>\n\n

Example<\/strong>: A business initially uses linear regression to predict customer lifetime value but later adopts a machine learning model to account for more complex interactions, resulting in more accurate predictions.<\/p>\n"}],"behaviour":{"enableRetry":true,"enableSolutionsButton":true,"slack":10,"feedbackMode":"onRequest"},"l10n":{"copy":"Copy","checkAnswer":"Check","tryAgain":"Retry","showSolution":"Show solution","allSectionsGood":"Your sections' lengths are all fine.","sectionTooShort":"Your section "@title" is too short. You need at least @chars more characters.","sectionTooLong":"Your section "@title" is too long. Remove at least @chars characters.","tooShort":"@chars characters too short","tooLong":"@chars characters too long","copyToClipboardSuccess":"Your text was copied to the clipboard","copyToClipboardError":"Your text could not be copied to the clipboard","section":"Section","messageNoSection":"There was no section given for this structure strip."},"a11y":{"feedback":"Feedback","copyToClipboard":"Copy text to clipboard","closeWindow":"Close window","showHints":"Show hints"},"taskDescription":"

Once a model has been evaluated, the next step is refinement. Refining a model means adjusting parameters, incorporating new data, or even switching to a different modeling technique. Here\u2019s how businesses can refine their predictive models:<\/strong><\/h3>\n"}

Refining Predictive Models

Rajbir Kaur