Adventure awaits: Elevating experiences with smart adventure recommenders

Tailoring outdoor experiences involves challenges like tracking, UI design, and feedback amalgamation. AI, Big Data offer exciting research prospects.

Adventure tourism is a popular choice for energetic individuals. Countries that attract tourists offer various adventure sports activities, which can be grouped into three main types: land, water, and air-based adventures. Furthermore, based on the risk level, they can be further classified into two main categories: hard and soft adventure activities. Soft adventures are suitable for those who prefer a safe experience, while hard adventures are more challenging and risky. People of all ages enjoy land-based activities like hiking, climbing, and running. As these activities gain popularity, there is a growing need for software solutions that will assist hikers, runners, and climbers in choosing the right routes or paths to follow.

Recommender systems

Recommender systems are tools designed to assist decision-makers, aiming to address specific problems. With the widespread use of smartphones, software companies have invested in creating various apps for outdoor sportsmen. However, these apps typically focus on a limited range of adventure sports activities, such as hiking, running, and climbing. This situation underscores the need to establish a standardised approach for building recommender systems in the adventure tourism sector.

To meet this need, we present a unified framework for interested parties, primarily software companies looking to offer guidance to their users in the realm of land-based adventure tourism, specifically hiking, running, and climbing. This framework aims to answer the key question: “How can someone recommend items related to land-based adventure tourism sports activities?”

Tailoring outdoor experiences involves challenges like tracking, UI design, and feedback amalgamation. AI, Big Data offer exciting research prospects.
Credit. Midjourney

Main findings

To answer the main question, we’ve identified several items that can be recommended in the adventure tourism field: routes, sequences of routes, destinations, training, management, diet, health, injury prevention, footwear, music, and virtual coaching. Additionally, there are six main methods for recommender systems that can guide stakeholders on these different items: content-based, collaborative filtering, knowledge-based, community-based, utility-based, and hybrid approaches.

It’s crucial to take the user’s context into account when providing relevant suggestions, known as context-aware recommendations. This context can be determined through three main techniques: pre-filtering, post-filtering, and contextual modelling. Furthermore, the context should include, among other things, weather information, the season of the year, the time of day, and the user’s location.

User profiling

Another critical aspect to consider is user profiling, which involves understanding users’ preferences and potential choices. User profiles can be created in two ways: through explicit feedback (where users manually provide their preferences) or implicit feedback (where the system models user preferences without direct input). The characteristics to be considered in a user’s profile are as follows: their physical abilities and limitations, training objectives, training achievements, physical attributes during specific activities, dietary preferences, perception of risk, motivation, expected gains, social preferences, and preferences related to elevation gain and steepness.

Items profiling

Item profiling is the process of examining the characteristics of items. These characteristics can be divided into two categories: personalised (which can vary depending on who the item is recommended to) and non-personalized. Personalised features include diversity, estimated time based on individual preferences, risk level, uniqueness and attractiveness of the item, difficulty, beauty, and overall appeal. On the other hand, non-personalized characteristics encompass distance, height, natural surroundings, solitude level, elevation gain/steepness, presence of natural light, proximity to nature, distance from traffic roads, type of terrain, air quality, and noise pollution.

Software functionality

In the world of software solutions, just like in other domains, the functionality of recommender system apps is a vital factor in their success. These apps should take into account essential aspects of software applications, including user satisfaction, user needs, app features, statistics, and the value they add. Recommender systems can be accessed through various output devices, such as smartwatches, smartphones, web browsers, and other communication devices.

Recommender system evaluation

Finally, an important aspect is evaluating recommender systems. This involves both offline and online assessments. Offline evaluation relies on measuring the error between predicted and actual values using several metrics like root mean squared error and mean absolute error. In contrast, online evaluation involves real users participating in experiments and providing feedback in response to specific questions. Three types of studies can be conducted in online evaluation: formative, user, and field studies.


From our comprehensive review of over a thousand scientific publications, it’s evident that research on recommender systems in adventure tourism is still relatively young. Currently, only three sports activities, namely hiking, running, and climbing, have received significant attention. Furthermore, running has more published papers compared to hiking and climbing. Many other adventure tourism activities, such as paragliding, sailing, and kayaking, remain largely unexplored in the research literature.

Future challenges

There are some limitations to the items recommended by these systems. Most studies focus on routes, paths, and training, while fewer explore other aspects like diet, injury prevention, sequential recommendations, and clothing choices. Similarly, when it comes to the methods used by recommender systems, many studies lean towards content-based and knowledge-based approaches. Only a small portion of the articles consider contextual information, which is an area worth exploring further. Social and hybrid methods are rarely used (found in only 2% of documents each), and there is still a need for more widespread use of utility-based methods.

Compared to recommender systems in non-sports domains like TV, music, or books, those in adventure tourism are less developed. However, with the advancements in artificial intelligence technologies and the availability of Big Data, there is now an opportunity to delve deeper into this area. While this work discusses many interesting systems and their advantages and limitations, there are still challenges that need to be addressed, including:

  1. Unobtrusive tracking technology and how it can be used for user profiling.
  2. The interface of the recommender system.
  3. There is a need to combine implicit and explicit feedback for user profiling, as there is no clearly defined methodology for this purpose.

All those points form an excellent direction that research should follow in the future.


Journal reference

Ivanova, I., & Wald, M. (2023). Recommender Systems for Outdoor Adventure Tourism Sports: Hiking, Running and Climbing. Human-Centric Intelligent Systems3(3), 344-365. https://doi.org/10.1080/13284207.2022.2155034

Iustina Ivanova is a computer scientist with a background in software engineering, and she is currently applying Artificial Intelligence in real-world applications. She completed her Master's in Artificial Intelligence and her Bachelor's in Software Engineering. Iustina have always been fascinated by computer vision and its application in software solutions. In 2019, she started a PhD program in computer science, and within the past three years, she has published a couple of noteworthy papers on computer vision solutions and recommender systems for sport climbers. Nowadays, she is concentrating on vision-based methods and their practical applications, where Artificial Intelligence can be utilised to enhance the quality of life.