Miley Posted May 2, 2018 Report Share Posted May 2, 2018 So I have been thinking of using correlation in my math IA. I want to look at the relationship between the amount of likes on a picture on instragram and the time it has been posted using datas from celebrities. Do you think this would work? and do you think I can make other analytical exploration about it? Reply Link to post Share on other sites More sharing options...
Drake Glau Posted May 2, 2018 Report Share Posted May 2, 2018 I'm seeing 1 pretty big hole that you could actually fill if you want but... Let's say a picture is posted at 1am and then the numbers of likes you record is after a few days where that 1am mark has passed multiple times. At this point you're not really recording data for when it was posted since you don't know which 1am mark you're recording from. I'm not doing a good job explaining that. Unrelated to correlation however and maybe outside the scope of SL Math is vector machines which are used with some training data in the form of 1-2 variables and then whether or not the person is a celebrity (obviously the definition of celebrity will need to be defined by you in some way). Then you can create a model, in this case, that would try and predict whether a person is a celebrity based off their number of likes they received on their pictures. Another thing I'm thinking of is making sure you restrict the pictures you look at for being less than x hours old and then record the time it was posted and the likes it received in those x hours and you can try to find a correlation of likes vs hour posted for multiple different celebrities and then explore past that. Did those celebrities post during their primetime? i.e do they post during the day or in the middle of the night? is that causing differences in the likes they received? Is instagram more popular in the US (for example) which you would see from maybe a Chinese celebrity posting in the middle of the night which would be daytime for US so maybe it gets more likes because of that? Reply Link to post Share on other sites More sharing options...
Miley Posted May 3, 2018 Author Report Share Posted May 3, 2018 Thank you for your ideas it really helps!! This is about correlation. Do you think I can use the Pearson Coefficient Rule and figure out the correlation? Also, I am planning to use diffrentiation to get the rate of the change of likes and use that to see if there is any relationship between time and the rate of change of likes. Reply Link to post Share on other sites More sharing options...
Drake Glau Posted May 4, 2018 Report Share Posted May 4, 2018 Time passed since posting or time of day? If you want to differentiate on time of day you'll need to restrict samples to 24 hour time periods since it's going to get weird when time of day loops back to midnight. But I think you could find a correlation between the rate of change in likes and the time of day. Pearson's should be good enough to attempt drawing a conclusion of whether or not the current time of day has an affect of the number of likes a person would receive during that specific time period. I would attempt to find as many data points as possible during that 24 hour period (data points being # of likes and time of day in 24h format) since I believe you have a minimum number of data points to achieve for pearson's to be "valid." But if instagram holds that kind of continuous stats you should be pretty golden. Reply Link to post Share on other sites More sharing options...
Miley Posted May 6, 2018 Author Report Share Posted May 6, 2018 Thanks it helped so much!!! Reply Link to post Share on other sites More sharing options...
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.