Machine Learning for spending prediction

I got this idea a couple minutes ago, while shitting in the toilet. Usually it would be something like ShowerThought, but let’s call this one as ShittingThought

Better idea, let’s make this as a regular thing, let’s make this as a Tag 😂

So as you may… or may not know, I’ve been making this project that I call as #Project_Riwallet the past few…

Hmmm… I didn’t have any progress the past few weeks… LET’S PRETEND NO ONE SEE THAT

But for all seriousness, this further make me feel a proposal and prototype as really important. The main reason I haven’t had any progress on this project, and Letisha is the lack of lead

And, as you may also know, I’ve been learning Machine Learning the past few weeks as well

Most of that learning is for Project_Rhea, but the longer I learn machine learning, the more possibilities that I realize

While shitting in the toilet, for some reason (That I personally have no idea about), I was thinking about Riwallet (Human mind, and especially my thought, is a really weird thing)

And I thought to myself, what data that I would have to collect for Riwallet to work properly

Oh yeah… now I remember why… I was watching one of Graham Stephan vid, while I was looking at the Robinhood’s website, I saw their design, and got inspiration for Riwallet design

I guess my mind is not that weird after all

For backstory, this is a screenshot of the current state of Riwallet. The feature that I anticipated the most while programming and planning, also that I proud the most of, is that one little graph on top of the app

Project Riwallet Early Screenshot

Ironically, I will probably replace it soon enough

Butt… now that I aware of the capabilities of Machine Learning, some of it at least, I thought wouldn’t it be cool if I can predict my future spending

And eaps, that is indeed a good idea

Then continuing that idea, I thought what data would I need to predict future spending

First, I would definitely need that one individual user data in the past to predict his future spending

Like, by knowing his usual spending on certain day, certain week, certain date and certain month. I would predict his spending

LINEAR REGRESSION TO THE RESCUE! (Disclaimer, I have no idea what I’m talking about, but it sound like a cool sentence, so I just put it here, you know, it sounds cool. And by my current knowledge, I think Regression is not for prediction purpose, wkwkw)

By knowing day, I would know the probability of the user to spend more money on Sunday for example

By knowing week, I would know the probability of the user spending more money earlier in the month for example

By knowing month, I would know the probability of the user spending in respect to current season, like maybe user will spend more for warm clothing in winter

And date, I think to compare year by year, idk, maybe like week. Like, maybe the user spend money the day they got their salary or something

Second, I would definitely need general data. Like, according to demographic, age range, region, and maybe global crisis (Like US-China trade war, damn I hate this)


Then after I gave myself some more time to think about it, I realize this prediction would be useless


Why? The factor that would render this prediction useless is, by knowing the prediction, people would feel challange by this prediction, maybe something like, huh, you think I would spend Rp. 40000 tomorrow? Let’s see about it

At least, that is what I would feel. You know, game theory, or something cool sounding like that

My hope is, people would spend less than the prediction, that would be cool

Also, I just rethink about that date data again, I think that would be useful to predict spending according to regular payment or spending, like, in my case I (hope) spend some money every certain date for my Reksa Dana or something like that, maybe paying rent like electricity and internet, or in my case, server

So yeah,

I think explaining feature that I haven’t implement is not a good idea since other people can steal the idea from me, but well, looking at this blog stats, I wouldn’t need to worry about that

So yeah, I really need to focus on one project at a time~

Leave a Reply

Your email address will not be published. Required fields are marked *