Non-Personalized Methods

Motivating Non-Personalized Recommenders

• In general, non-personalized recommendations are usually generated for an average user
• Since the average profile is associated with such a small percentage of users, these non-personalized recommendations aren't very useful
• Non-personalized recommendations aren't usually accurate compared to personalized recommendations
• However, they can be useful when we don't have information about certain users (or used in hybrids)

Defining Types of Non-Personalized Recommendations

• Popular Items: popular categories, brands or products
• New Releases: new products released
• Similar Items: similarly purchased products

• Trending Items: products trending upwards

  • Where, $s_{1}$ is the percentage change in sales for the previous day
  • And, $s_{2}$ is the percentage change in sales two days ago

References

• Textbook about Algorithmic Marketing