Not necessarily, just because you never ordered it doesn’t mean it’s not a potential recommendation. Actually it may be more likely they will recommend things you haven’t ordered before as the system is trying to get you to potentially have more diversity of products. Even in the example above, yeah a dip looks stupid from just the recommendations, but what if the thing being ordered is bread sticks or wings, then it’s normal.
If you offer “recommended dishes”, you should probably be accounting for different kinds of food items to enable that. I wouldn’t say the pure machine learning aspect is necessarily directly at fault, but the end to end implementation is bad.
Yeah, of course that doesn’t mean it’s good machine learning though 😆
Not necessarily, just because you never ordered it doesn’t mean it’s not a potential recommendation. Actually it may be more likely they will recommend things you haven’t ordered before as the system is trying to get you to potentially have more diversity of products. Even in the example above, yeah a dip looks stupid from just the recommendations, but what if the thing being ordered is bread sticks or wings, then it’s normal.
If you offer “recommended dishes”, you should probably be accounting for different kinds of food items to enable that. I wouldn’t say the pure machine learning aspect is necessarily directly at fault, but the end to end implementation is bad.