Types of Problems That Machine Learning Can Solve

Regression: Estimating a continuous numerical variable from a set of input variables. For example, predicting the purchase price of a house based on its features, or predicting the performance of a product that is still in design phase.
Classification: Assigning data records to predefined categories or classes of two values (binary classification) such as spam/no-spam in an email application, or more than two values (multi-classification) such as minor/major/critical in categorizing the severity of a fault based on the shared characteristics of the input features.
Clustering: Exploring data and finding natural groupings based on similarity and other measures of structure inherent in the data. A common application is creating customer segments for enhanced targeting based on data about the demographics, preferences and buyer behavior of existing customers.
Associations: Also known as “affinity analysis” and “market basket analysis”, used in finding combinations of items, conditions, or other attributes that tend to occur together in a single event or transaction. Examples of applications are increasing sales by identifying products that should be stocked together on the store shelves, and reducing insurance claims fraud by understanding claim details, claim history, payment history, and client attributes commonly found in fraudulent claims.
Sequences: Finding combinations of items, conditions, or other attributes that tend to occur together in a series of events or transactions. Use cases include identifying the service/product a customer will most likely buy after purchasing a new house and new furniture; and identifying the part that is most likely to fail given a machine’s repair history.
Ranking: Identifying the status or position of an item on a scale. This is commonly used in presenting internet search results, advertising, and to identify factors that drive certain behaviors such as churn or fraud.
Optimization: Finding the best solution from a range of choices available in a certain mostly complex situation. Common applications are minimizing cost, maximizing profit, and finding the best store to deliver a product relative to a distribution center.
Anomaly detection: Identifying unusual or suspicious cases based on deviation from the norm. Common applications include detecting fraud in insurance claims, detection issues in tax compliance, and early identification of problems in equipment for predictive maintenance.
Forecasting: Predicting future developments in business based on analysis of trends in time series data. Usage examples include estimating expenditure, predicting weekly sales to avoid the loss of perishable goods, and predicting whether a customer will churn in the next 3 months.
Recommendation: Suggesting the most relevant item, mostly products, based on historical data. Common examples are suggested articles on online newspapers, video recommendations on YouTube or Netflix, and connection suggestion on LinkedIn.
Generation: Creating an audio, video, image or text from given input data. Applications include audio transcription, language translation, and image recognition.

  • Join WhatsApp group here
  • Join Facebook group here
  • Follow on LinkedIn here

Powered by BetterDocs

Leave a Reply

Your email address will not be published. Required fields are marked *

Scroll to top