A Report on an Expert System for Netflix Recommendations During Quarantine

A Report on an Expert System for Netflix Recommendations During Quarantine

Introduction

With the current state of the world being where it is due to the recent outbreak of the COVID-19 virus, millions of people around the world have been isolating themselves indoors, be it willingly or forcefully by law. With this, a good majority of those isolating indoors share one thing in common: they’re bored out of their minds. People are looking for anything to entertain themselves to help pass the time, as this COVID-19  pandemic does not seem to be slowing down at the time of writing this report. While some love to be indoors, many are struggling to maintain indoors for this extended duration of time. Those that love to be outdoors are now restricted to participate in some of their favorite hobbies: going to the park, going out to eat, meeting up with friends, etc. There are plenty of outdoor hobbies that just cannot be replicated indoors, for instance, you could attempt to recreate your favorite meal from a restaurant, but deep down you will not feel the same consuming that meal. But, there is one hobby that has not been hindered, and in a way has actually become a saving grace in these trying times to find entertainment: watching movies.

Although we have been stripped of the privilege to go visit our local cinema and watch one of the newest releases, we have been blessed with current-day streaming services such as Disney+, Amazon Prime Video, and the most popular, Netflix. Netflix, founded in 1997 by Reed Hastings and Marc Randolph, started as an online DVD rental website. In 1999, Netflix moved to a subscription service style, allowing users to rent unlimited DVDs all for one low monthly price.  Popularity grew and so did the total subscribers to Netflix in the US. In 2007, Netflix implemented its greatest innovation, the option to users to stream television shows and movies from their personal computer. From then it was history, as technology advanced and more personal devices were created, Netflix was made available on them, from gaming consoles, to the tiny phone in your pocket (“About Netflix”, 2020). Fast-forward to January of 2020, and it is reported that Netflix has an active subscriber count of 167 million (Clark, 2020). And breaking it down further, based on a Streaming Observing calculation of the average daily viewing of Netflix being 71 minutes, that comes out to roughly 165 million hours of  Netflix watched daily across the globe. With that being said, it is undeniable of just how popular Netflix is, and to some, how important Netflix is to them, especially during a time like this.

As young students, Netflix has played a huge impact and has become the primary supplier of media for most of us. So when the opportunity presented itself for us to make a small expert system that looked to solve a problem, we felt that the problem of “What should I watch on Netflix?” was of utmost relevance and importance, perhaps now more than ever. Although Netflix has been a blessing by providing an ease of access to a wide array of movies and tv shows, it has been accompanied with a slight downside due to this same wide array, it is usually difficult to select a sole option to watch. Which is where our inspiration for this expert system came, as a group, with the self-isolation in effect, we have been utilising a google chrome extension by the name of “Netflix Party” in order to watch netflix simultaneously with each other. It has been quite a pleasant experience but before we can enjoy a film, we always seem to spend countless times trying to decide what we should watch. So the idea of an expert system that can recommend a movie or tv-show available on Netflix was an actual problem we personally face, so why not look to resolve it. We believe that watching Netflix may not be in the utmost importance to people during these times, seeing as there are plenty more things that should be prioritized, we also believe finding a form of entertainment to keep ourselves preoccupied is quite important to ease the pain of having to stay indoors, and outright improves mental health.

The only thing left to consider was the potential of other similar expert systems that are currently available, to which the answer is that there are others available. There are a couple Netflix recommendation systems available online, but the best one we found was from whatthehellshouldiwatchonnetflix.com. This website’s system is very similar to that of our own whereas you will go through the system question-by-question until the system deems fit to output a recommendation. The downside of this website is that it is limited to two pages of questions to which you are first asked to choose from a movie, tv show, or random, followed by the choice of genre. Which is why the system is quite underwhelming as it does not produce a narrowed down answer. But what’s great about the system is the personalised reviews of each film and tv show, and also the option to just outright randomize from the get go. Other than that, many of the other available systems are not a series of question models, whereas you would be required to answer one question before being able to answer the next. But instead, many of these sites apart from whatthehellshouldiwatchonnetflix.com, have the user input their prerequisites all at once and then output an answer based on that. But apart from all of these systems available, there is one that we cannot go without mentioning, the integrated Netflix Recommendation system.

Within the Netflix application, there is an integrated recommendation system that is taking in information from users constantly to best recommend shows and movies. Netflix details that the system estimates a show or movie that you are most likely to watch based on factors such as viewing history, ratings of similar titles, time of day you watch, device watched on, and length of time watched. The system also utilises the information from other users that exhibit the same qualities as the user to be recommended as it is more likely they would like the same shows and movies. The Netflix application also has a “jump start” to their recommendation as upon launching the application for the first time, it will ask the user to choose from select titles to use as an initial benchmark for recommendations. The rows within Netflix are also ranked using algorithms, and delving deeper, with each row, titles are once again ranked; highest recommendation to the left, and further down the row is lower recommendations (“About Netflix”, 2020).

Knowledge Engineering

Knowledge engineering refers to aspects involved in creating knowledge-based systems. These aspects include those of technical, scientific, and social degrees in order to best build and maintain said systems (“Knowledge engineering”, 2020). In the case of our own expert system that we created, it is necessary that we utilize the information of an expert to best provide a suitable output to answer the question of “What should I watch on Netflix?”. In the case of recommending an option to view on Netflix, the type of expert that would be most capable of answering that question would be that of someone who critiques tv shows and movies. The only difficulty with utilizing a sole critic’s reviews to create our expert system based on recommending something on Netflix brings us back to the problem that derived our reason to create said expert system, there’s an enormous catalog of options available on Netflix that no single person could ever watch entirely. To put it into perspective, in the year of 2019 alone, Netflix produced and released 371 Netflix original series and movies, that’s more than one series/movie per day (Elliot, 2019). That’s limiting just to Netflix originals, looking at the entire catalogue available to Netflix Canada subscribers shows that there are currently 5914 different series and movies at the time of writing this report (“Search The Full Canadian Catalogue”, 2020). To put it simply, it is quite literally impossible for one sole person to view everything available on Netflix. But then, how could we compile enough “expert” information to create our system?

Our primary sources of information gathered to create this expert system are from two places: Netflix itself, and reviewer rating sites such as Metacritic. Starting off with Netflix itself, we have previously discussed in this report the integrated recommendation system available in the application. Upon viewing our own individual Netflix accounts and options that are recommended to us, we were able to get a vague understanding of categorizing and grouping films and series that may be of interest to a user based on their previous viewing history. This helped in best dividing our output options to their dedicated response sub-group such as genre, MPA rating, etc. From there, the most crucial source of “expert” information would come from review rating sites such as Metacritic.

With the previous point stated being that no mere individual could ever watch the entire Netflix catalogue by themself, the best option in order to get expert information on said series and tv shows would be to observe reviews from established critics. This is made possible by websites such as Metacritic. Metacritic, founded in 1999, is a review aggregator site that provides excerpts of critic reviews from multiple sources and compiles a weighted  average score based on said reviews to create a cumulative score to rate films, tv shows, music albums, and video games (“Metacritic”, 2020). With access to a source such as Metacritic, an individual would not have to rely on a sole “expert”, in this case a critic, but instead can receive feedback from a qualified individual who has seen the in particular film or series they are interested in. This access allowed us to provide qualified expert knowledge pertaining to the recommendation of the titles that we have included in our system.

Methodology

Technical Overview

Benefits

When it comes to the benefits of using our expert system, the two main benefits would have to be ease of use, and the inclusion of taking account for miscellaneous factors. Starting with the ease of use, our system is very straightforward. The questions provided to gather information are all very simple, and should not have the user be confused nor pondering the answer for very much time apart from the final question before output whereas the user would select from the different possible genres. It is understandable that the user may take some time answering this in particular questions but otherwise, the questions preceding are of low thought process. As for the benefit of taking account miscellaneous factors, this pertains to questions that are near the beginning of our expert system. The factors to mention consist of: type of device used, time of day viewing, and time spent viewing. To some these questions may be unnecessary to answer a question like “What should I watch on Netflix?”, but in the grand scheme of things, these factors can be the difference between an alright recommendation and an outstanding one. For instance, a user looking to watch a film during the day but also wants an action film would be best suited for a more light hearted action film, opposed to a user looking for an action film at the end of their day, a climactic action film would be best suited to conclude their day.

When speaking about the strengths of our system, it is also good to play devil’s advocate and look at some of the weaknesses. The major con of our expert system would have to be due to technical applications. With the system being run based off of hyperlinks created within individual word documents, and added to the actual experience each member of the group has had with coding, it is understandable that our end product is not of market ready quality. We are simply a couple of business students who wanted to create an “expert” system relevant to us, so having the system not be capable of absolutely recommending any possible option available on Netflix should be understandable. Otherwise the system is capable of completing the task it was meant to do, just with some limitations. Which leads to where we would go next with our system if we were to continue development of it.

Next Steps

If the possibility of continuing the development of our expert system presented itself, it is quite simple as to what direction we would look to take. Firstly, we would develop a more presentable system, rather than utilising the likes of hyperlinks in a Microsoft Word document, but instead maybe javascript. The website would be overhauled with graphics and other details to create a more presentable initial boot-up of the system. Secondly, we would aim to continue to expand the possible outcomes of recommendations, this can also be accompanied with the inevitable implementation of additional questions to create a deeper input of information to create a greater output recommendation. With more potential information inputted, there would be potential for a far greater amount of results, leading all the way to the final stage of the system being that it would accommodate for every possible option available on Netflix. Lastly, with the slow implementation of greater possible outcomes, we would look to develop the final result screen displaying the resulting recommendation. We would look to implement something similar to that of whatthehellshouldiwatchonnetflix.com, whereas the results would tend to have a personalised review/overview of the recommendation. Another great integration would be that of direct link to Metacritic reviews along with the Metacritic score being displayed with the recommendation itself.

Conclusions

Works Cited

About Netflix. (n.d.). Retrieved from https://media.netflix.com/en/about-netflix

Clark, T. (2020, January 24). Netflix is still growing wildly, but its market share has fallen to an estimated 19% as new competitors emerge. Retrieved from https://www.businessinsider.com/netflix-market-share-of-global-streaming-subscribers-dropping-ampere-2020-1

Elliott, M. (2019, December 25). You Won’t Believe How Many Original Movies and Shows Netflix Released in 2019. Retrieved from https://www.cheatsheet.com/entertainment/you-wont-believe-how-many-original-movies-and-shows-netflix-released-in-2019.html/

Knowledge engineering. (2020, March 19). Retrieved from https://en.wikipedia.org/wiki/Knowledge_engineering

MaFt.co.uk. (2020). Search The Full Canadian Catalogue. Retrieved from https://can.newonnetflix.info/catalogue

Metacritic. (2020, April 1). Retrieved from https://en.wikipedia.org/wiki/Metacritic

Netflix Revenue and Usage Statistics (2020). (2020, March 6). Retrieved from https://www.businessofapps.com/data/netflix-statistics/