Strange Recommendations? On the Weaknesses of Current Recommendation Engines
Currently, most approaches to recommendation engines focus on traditional techniques such as collaborative filtering, basket analysis, and content-based recommendations. Recommendations are considered from a prediction point of view only, i.e., the recommendation task is reduced to the prediction of content that the user is going to select with highest probability anyway. In contrast, in this chapter we propose to view recommendations as control-theoretic problem by investigating the interaction of analysis and action. The corresponding mathematical framework is developed in the next chapters of the book.
KeywordsUser Behavior Recommendation Algorithm Good Recommendation Profitable Product Recommendation Engine
2.1 Introduction to Recommendation Engines
Recommendation engines (REs) for customized recommendations have become indispensable components of modern web shops. REs offer the users additional content so as to better satisfy their demands and provide additional buying appeals.
There are different kinds of recommendations that can be placed in different areas of the web shop. “Classical” recommendations typically appear on product pages. Visiting an instance of the latter, one is offered additional products that are suited to the current one, mostly appearing below captions like “Customers who bought this item also bought” or “You might also like.” Since it mainly respects the currently viewed product, we shall refer to this kind of recommendation, made popular by Amazon, as product recommendation. Other types of recommendations are those that are adapted to the user’s buying behavior and are presented in a separate area as, e.g., “My Shop,” or on the start page after the user has been recognized. These provide the user with general but personalized suggestions with respect to the shop’s product range. Hence, we call them personalized recommendations.
Further recommendations may, e.g., appear on category pages (best recommendations for the category), be displayed for search queries (search recommendations), and so on. Not only products but also categories, banners, catalogs, authors (in book shops), etc., may be recommended. Even more, as an ultimate goal, recommendation engineering aims at a total personalization of the online shop, which includes personalized navigation, advertisements, prices, mails, and text messages. The amount of prospects is seemingly inexhaustible. For the sake of simplicity, however, this book will be restricted to mere product recommendations – we shall see how complex even this task is.
Recommendation engineering is a vivid field of ongoing research. Hundreds of researchers, predominantly from the USA, are tirelessly devising new theories and methods for the development of improved recommendation algorithms. Why, after all?
Of course, generating intuitively sensible recommendations is not much of a challenge. To this end, it suffices to recommend top sellers of the category of the currently viewed product. The main goal of a recommendation engine, however, is an increase of the web shop’s revenue (or profit, sales numbers, etc.). Thus, the actual challenge consists in recommending products that the user actually visits and buys, while, at the same time, preventing down-selling effects, so that the recommendations do not simply stimulate buying substitute products and, therefore, in the worst case, even lower the shop’s revenue.
This brief outline already gives a glimpse at the complexity of the task. It is even worse: many web shops, especially those of mail-order companies (let alone bookshops), by now have hundreds of thousands, even millions, of different products on offer. From this giant amount, we then need to pick the right ones to recommend! Furthermore, through frequent special offers, changes of the assortment as well as – especially in the area of fashion – prices are becoming more and more frequent. This gives rise to the situation that good recommendations become outdated soon after they have been learned. A good recommendation engine should hence be in a position to learn in a highly dynamic fashion. We have thus reached the main topic of the book – adaptive behavior.
We abstain from providing a comprehensive exposition of the various approaches to and types of methods for recommendation engines here and refer to the corresponding literature, e.g., [BS10, JZFF10, RRSK11]. Instead, we shall focus on the crucial weakness of almost all hitherto existing approaches, namely, the lack of a control theoretical foundation, and devise a way to surmount it.
2.2 Weaknesses of Current Recommendation Engines and How to Overcome Them
If the products (or other content) proposed to a user are those which other users with a comparable profile in a comparable state have chosen, then those are the best recommendations.
Or in other words:
Approach I: What is recommended is statistically what a user would very probably have chosen in any case, even without recommendations.
This reduces the subject of recommendations to a statistical analysis and modeling of user behavior. We know from classic cross-selling techniques that this approach works well in practice.
The effect of the recommendations is not taken into account: If the user would probably go to a new product anyway, why should it be recommended at all? Wouldn’t it make more sense to recommend products whose recommendation is most likely to change user behavior?
Recommendations are self-reinforcing: If only the previously “best” recommendations are ever displayed, they can become self-reinforcing, even if better alternatives may now exist. Shouldn’t new recommendations be tried out as well?
User behavior changes: Even if previous user behavior has been perfectly modeled, the question remains as to what will happen if user behavior suddenly changes. This is by no means unusual. In web shops, data often changes on a daily basis: product assortments are changed, heavily discounted special offers are introduced, etc. Would it not be better if the recommendation engine were to learn continually and adapt flexibly to the new user behavior?
Optimization across all subsequent steps: Rather than only offering the user what the recommendation engine considers to be the most profitable product in the next step, would it not be better to choose recommendations with a view to optimizing sales across the most probable sequence of all subsequent transactions? In other words, even to recommend a less profitable product in some cases, if that is the starting point for more profitable subsequent products? To take the long-term rather than the short-term view?
These points all lead us to the following conclusion, which we mentioned right at the start – while the conventional approach (Approach I) is based solely on the analysis of historical data, good recommendation engines should model the interplay of analysis and action:
Approach II: Recommendations should be based on the interplay of analysis and action.
In the next chapter, we will look at one such approach of control theory – reinforcement learning. First though we should return to the question of why the first approach still dominates current research.
Part of the problem is the limited number of test options and data sets. Adopting the second approach requires the algorithms to be integrated into realtime applications. This is because the effectiveness of recommendation algorithms cannot be fully analyzed on the basis of historical data, because the effect of the recommendations is largely unknown. In addition, even in public data sets, the recommendations that were actually made are not recorded (assuming recommendations were made at all). And even if recommendations had been recorded, they would mostly be the same for existing products because the recommendations would have been generated manually or using algorithms based on the first approach.
This trend was further reinforced by the Netflix competition [Net06]. The company Netflix offered a prize of 1 million dollars to any research team which could increase the prediction accuracy of the Netflix algorithm by 10 % using a given set of film ratings. The Netflix competition was undoubtedly a milestone in the development of recommendation systems, and its importance as a benchmark cannot be overstated. But it pushed the development of recommendation algorithms firmly in the direction of pure analytics methods based on the first approach.
So we can see that on practical grounds alone, the development of viable recommendation algorithms is very difficult for most researchers. However, the number of publications in the professional literature treating recommendations as a control problem and adopting the second approach has been on the increase for some time.
As a further boost to this way of thinking, prudsys AG chose the theme of recommendation algorithms for its 2011 Data Mining Cup, one of the world’s largest data mining competitions [DMC11]. The first task related to the classical problem of pure analysis, based however on transaction data for a web shop. But the second task looked at realtime analytics, asking participants to design a recommendation program capable of learning and acting in realtime via a defined interface. The fact that over 100 teams from 25 countries took part in the competition shows the level of interest in this area.
A further example of new realtime thinking is the RECLAB project of RichRelevance, another vendor of recommendation engines. Under the slogan “If you can’t bring the data to the code, bring the code to the data,” it offers researchers to submit their recommendation code to the lab. There, new algorithms can be tested in personalization applications on live retail sites.
- [BS10]Bhasker, B., Srikumar, K.: Recommender Systems in E-Commerce. Tata McGraw-Hill Education (2010)Google Scholar