IGCSE Computer Science 0478: Machine Learning vs Expert Systems (Fitness App)

A gym's mobile fitness app suggests a personalised workout to each member. After every session, a member records how many repetitions they completed and rates how difficult the session felt. The app uses this growing store of data, together with machine learning, so that the reps and difficulty it suggests next time automatically become better suited to that member, without a programmer ever rewriting the app's code.

(a) State what is meant by machine learning. [1]

(b) Describe how the fitness app's behaviour is an example of machine learning, rather than a fixed set of rules written once by a programmer. [3]

(c) The gym also uses an expert system to help staff choose which of three standard injury-recovery plans to recommend to a member, based on the member's answers to a fixed set of questions about their injury. Explain one difference between this expert system and the machine-learning fitness app in how each one changes, or does not change, over time. [2]

(d) State one disadvantage of using machine learning, rather than a fixed set of rules, to decide a member's suggested workout. [1]

Part (a): Definition of machine learning

Machine learning is when a program has the ability to automatically adapt its own processes and/or the data it uses, rather than needing a person to reprogram it.

Part (b): Why the fitness app is an example of machine learning

Each time a member completes a session, new data, reps completed and a difficulty rating, is added to what the app already knows about that member.
The app uses this accumulating data to automatically change the reps and difficulty it suggests for future sessions; its process adapts based on the data, without a programmer rewriting any fixed rules.
This is different from a fixed rule, such as “if the member is a beginner, always suggest 10 reps”, which would carry on giving the same suggestion forever regardless of how the member’s performance changes, because a fixed rule set does not adapt itself.
Because the app’s own suggestions change automatically as more data is gathered, this is machine learning rather than a fixed rule-based program.

Part (c): Expert system vs machine learning over time

The expert system’s knowledge base and rule base for recommending a recovery plan stay the same each time it is used. It will keep giving the same recommendation for the same answers unless a person manually updates its stored rules. The machine-learning fitness app, by contrast, automatically adapts its own suggestions as more member data is collected, without anyone rewriting it. The key difference is that the expert system only changes if a person edits it, whereas the machine-learning app changes itself.

Part (d): A disadvantage of machine learning here

Because the app’s suggestions change by themselves based on the data gathered, it can be difficult for staff, or the member, to know exactly why it changed a workout. If a member logs inaccurate or unusual data, for example a mistaken difficulty rating, the app may adapt its suggestions in an unhelpful or even unsafe direction, without a programmer’s fixed rules to prevent this.

Final answers

(a) Machine learning is a program automatically adapting its own processes and/or data.
(b) The app’s suggestions change by themselves as more session data builds up, rather than following one fixed, unchanging rule.
(c) The expert system only changes if a person edits its knowledge/rule base; the machine-learning app adapts itself automatically.
(d) Any suitable disadvantage, e.g. suggestions can change unpredictably or be hard to explain, and inaccurate logged data can push the app’s suggestions in an unhelpful direction.