Study the past to know the future – Confucius. While building product recommenders, the cold start is one of the biggest challenges that developers face, in addition to others such as data scarcity, scalability, and diversity. Read this blog to know more about Machine Learning Solutions for the Cold Start Problem in Recommender Systems.

Synonymous with a car engine failing to start in cold weather, the cold start problem confounds recommenders across two categories: product and user. When a new product or service hits the market, it has no likes, dislikes, reviews, or clicks, making it difficult to recommend.

It's the same when a new user signs up: since the system has no information about them or their preferences, it can't recommend a product. It's like a clean slate, so the opportunity of studying the past to forecast the future simply does not exist.

Want to use our recommender system to improve conversions? Get in touch with our experts

Understanding the Cold Start Problem

One of the most popular methods used by classic recommender systems is collaborative filtering (CF). The latter presumes that the user or the item has some ratings, from which it can then infer the ratings of similar users or items.

But for new users or items, the recommender fumbles because of a lack of associated data, so one cannot draw assumptions using the standard matrix factorization technique, a collaborative filtering method to find the correlation between items and users.

So, collaborative filtering cannot be used for cold starts because it relies on user actions to recommend items (products advertised on your site). A CF model surfaces more and more items with increasing accuracy as a user continues to interact with the website or app over time.

But at the start line, i.e., cold start, since the user (or the product) is new, there's no track record to fall back on.

In today's world of commercial artificial intelligence (AI), the cold start is also an opportunity for the solution to include a degree of automated data modeling.

In their research paper, "Treating Cold Start in Product Search by Priors", researchers Parth Gupta, Tommaso Dreossi, Jan Bakus, Yu-Hsiang Lin, and Vamsi Salaka explain the problem explicitly:

Learning to Rank (LTR) models rely on several features to rank documents for a given query. Many LTR features are based on users' interactions with documents, such as impressions, clicks, and purchases. We call these features behavioral features.

Ranking models are trained to optimize user engagement, and therefore, such behavioral features tend to be the most important training signals.

However, new and tail products that do not have user engagement lack behavioral features and hence are ranked as irrelevant, which in turn further excludes them from catching user engagement. It takes time for them to gather enough behavioral signals to show up at their fair ranking position.

This leads to the causality dilemma: No behavioral data causes poor ranking, which in turn reduces the likelihood that new products will accrue behavioral data.

This phenomenon is referred to as a cold start problem and poses serious concerns, from bad customer experience to lost revenue opportunities.

Over the years, recommender systems have used various approaches to "kick-start" their recommendations in a cold-start setting. The early days' recommenders drew on the methods and theories taken from other artificial intelligence (AI) fields for user profiling and preference discovery.

But, of late, with the commercial deployment of AI, we have seen an increase in the success of AI apps.

Solutions for the Cold Start Problem in Recommender Systems

Developers have started deploying various AI techniques in recommender systems, as AI offers a higher degree of recommendation accuracy than conventional methods.

This has ushered in a new generation of recommender systems that combine data analytics and AI to generate advanced insights into the relationships between users and items.

In ML, for supervised learning, access to and the application of labeled data from the past are necessary to label previously unseen data for the future.

The same is true in the case of unsupervised machine learning, where patterns are characterized because no assumptions can be made regarding the existence of labeled training data.

Transform your business using Express Analytics' machine learning solutions >>>> Let's connect

So, then, the obvious question – how does one go about tagging labels for future data when there are no labels (diagnoses, classes, known outcomes) in the past data?

Here are some of the solutions that combine algorithms and the techniques of ML to produce a neutral judgment in model development and progression, which eventually leads towards optimization:

Content-based filtering

This seems to be among the favorite options used by ML developers for cold-start cases. The product recommender can use metadata about the new product to create recommendations.

What's more, any additional information, such as data obtained about the user from their social media networks (provided at the time of signing up with social logins), can also be used to address the initial information scarcity.

Also, here, filtering algorithms are given a predisposition toward specific items and therefore recommend similar items based on a domain-specific concept of item content.

There are many advantages to a content-based recommender system. To begin with, this type of recommendation is based on item representation, making it independent of the user. As a result, the issue of data sparsity does not arise.

Next, content-based recommender systems can suggest new items to users, thereby resolving the new-product cold-start problem.

Popularity-based model

Another go-to option frequently used for cold-start cases. Using Python to build the model, a new customer, at the very start of a customer journey, can be shown a list of trending (popular) products. Maybe, if there's a product on the list that almost all new customers buy, that, too, can be suggested to all new customers.

Following this, each choice can be registered along with contextual information, i.e., the location (using the device's GPS coordinates), the channel the visitor came from, the device used, and so on. After the first few clicks, behavioral science will help the e-commerce site build a customer profile to build up from there.

For a first-time product about to be listed, it has no previous "baggage", i.e., the purchase history, and so on. Till enough purchases or likes are gathered around this one product, eCommerce sites like Amazon and even YouTube continue to promote it frequently, in an almost in-your-face campaign.

What is also done simultaneously is to display similar products, using string similarity algorithms such as Levenshtein distance or Hamming distance, until enough users have either bought the product or rated the new service. There you have it, then, data around a product that proves its popularity or otherwise.

One drawback this model does have is a lack of personalization, but then again, you are talking of tackling the cold start problem in recommender systems, and not after the engine has been warmed and is running smoothly, figuratively speaking.

The multi-armed bandit model

The inspiration for this comes from a multi-level casino slot machine. The gambler has the option of pulling not one but many levers, so he thinks his chances at winning have gone up. (Incidentally, the single lever slot machine is called the single-armed bandit).

The probability ratio of the reward to each lever is different, though. The ratios are not known to the gambler (he is taking his chances) but known to the casino, which increases its profits eventually, because the House never loses (hence, the use of the term "bandit").

So, what relevance does this have to an algorithm that tries to circumvent the cold-start problem in recommender systems?

Every day, scores of new items arrive on a website or app. These items are like the various slot machines, single as well as multi-lever, and each comes with a different ROI. But because these are fresh items, the e-commerce site really does not know how many users will buy them.

In a multi-arm slot machine, each arm represents a reward. So, recommending a subset of the new products is like deciding to pull a subset of slot machines to draw. The trick for the site is to identify which lever to pull to eke out the maximum return.

There's an inherent problem with this kind of machine learning model, though. It's referred to as the "exploration vs exploitation" problem. In layman's terms, if a particular new product is flying off the shelves, it is but natural for the seller to get greedy and show it to many more users (called exploitation).

But at the same time, the remaining new products may be languishing, and the seller also needs to show them to customers at a sufficient frequency, because they may then turn out to be even more popular than the (popular) item the seller has already shown (exploration).

Clearly, a balance has to be struck between exploration and exploitation. There exist various bandit algorithms for making optimal recommendations for the user. The popular MAB algorithms include:

Epsilon Greedy

Decayed Epsilon Greedy

Upper Confidence Bound

Thompson Sampling

Each of these can be coded in Python. Eventually, the idea is to use any of these models so that the bandit algorithm helps the seller choose between exploiting the one item that gave it the highest reward and exploring the other products.

The seller has to learn to balance reward maximization based on the information already obtained against implementing new actions to increase the database. In ML, this is called the exploitation vs. exploration tradeoff.

Transform your business using Express Analytics' machine learning solutions >>>> Let's connect

Deep Learning Approach

With the recent advent of deep learning, a sub-discipline of machine learning that mimics the human brain, there are new attempts to resolve the cold start problem in recommender systems, as well as other issues, using this approach.

There are many models and research papers that suggest using deep neural networks to mitigate the cold-start problem in recommender systems. Using neural networks, the initial weights/ratings on the network edges are assigned randomly, but backpropagation is used to fine-tune the model to the optimal.

Meta-learning approaches

Meta-learning approaches have gained popularity in machine learning for learning representations that are useful across a wide range of tasks, according to this research. In his paper, "Meta-Learning for User Cold-Start Recommendation", Homanga Bharadwaj, Department of Computer Science and Engineering, Indian Institute of Technology Kanpur, India, says they designed a recommendation framework that was trained to be "reasonably good enough" for a wide range of users.

It was inspired by the generalizability of model-agnostic meta-learning. During testing, to adapt to a specific user, the model parameters were updated with a few gradient steps, and then the model was evaluated on three benchmark datasets.

The paper, using detailed simulation studies, shows that the framework handled the user cold-start model much better than state-of-the-art benchmark recommender systems.

Social network integration

Another research paper in the same field by a team of researchers from the University of KwaZulu-Natal, Westville, South Africa, proposed a new approach to solving the cold-start problem in recommender systems by leveraging social networks and Matrix Factorization to enhance a deep learning approach.

The research team explained that the social information was used to form user groups, since users within a given community were likely to share the same interests. For this, a community detection algorithm was used. Once they were segregated, a deep learning model was trained on each community.

The comparative models were then evaluated using Mean Squared Error (MSE) and Mean Absolute Error (MAE). The evaluation was carried out using 5-fold cross-validation.

The results showed that using social information improved the performance of the Deep Learning Approach, and that grouping users into communities was advantageous.

Advanced Solutions and Best Practices

Hybrid approaches

Combining multiple recommendation techniques can provide better results for cold start scenarios:

Content + Collaborative: Use content-based filtering for new items and collaborative filtering for established users

Popularity + Personalization: Start with popular items and gradually build personalization

Deep Learning + Traditional ML: Combine neural networks with classical algorithms

Feature engineering for cold start

User features: Demographics, location, device type, signup source

Item features: Category, price range, brand, release date

Contextual features: Time of day, season, current trends

Evaluation metrics for cold start

Click-through rate (CTR) for new items

Conversion rate for new users

Time to first interaction for new users

Diversity of recommendations

Implementation Considerations

Technical challenges

Scalability: Handling large numbers of new users/items simultaneously

Real-time processing: Making recommendations in milliseconds

Data quality: Ensuring clean data for new entities

A/B testing: Validating different cold start strategies

Business considerations

User experience: Balancing exploration with user satisfaction

Revenue optimization: Maximizing conversions while exploring

Long-term engagement: Building sustainable recommendation patterns

Privacy compliance: Handling new user data responsibly

Conclusion

The problem of cold start in a recommender system has been around for years, but with the advent of artificial intelligence coupled with data analytics, much progress has been made, and today, there are quite a few solutions around to overcome it.

The key is to choose the right approach based on your specific use case:

For new users: Start with popularity-based recommendations and gradually build personalization

For new items: Use content-based filtering and social proof

For both: Implement multi-armed bandit algorithms for optimal exploration-exploitation balance

For advanced scenarios: Consider deep learning and meta-learning approaches

The future of cold start solutions lies in more sophisticated AI techniques, better feature engineering, and hybrid approaches that combine the strengths of multiple algorithms.

Machine Learning Solutions for Cold Start Problem in Recommender Systems