# How to predict unknown(hidden) value by incomplete value or partly recorded value

Let me make it clear by make an example:

Suppose I knew a person's cost each month for 3 years like:

2016Jan : $2500 2016Feb :$4000

2016Mar : $3500 ... Just according to this, can I predict how much he earned each month in these years? If I could, what model should use? If I couldn't, what info I need to make this possible. If I couldn't guess in 100 percent but I could guess by some probability, what model should I use? Another example: Need to estimate how much each store sells in each city, however I could only got the sales when they record it in electric system, if they didn't record it in electric system(may use paper or something else); then I don't know, how can I get all sales, not just the sales record in electric system? Any suggestion is welcomed ## 1 Answer Actually you can use almost any regression model. When you do not want to go too much into theory simply use a gaussian process provided by sklearn. This model gives you predictions and uncertainties. • Thanks for let me know the gaussian process by sklearn – cloudscomputes Nov 14 '18 at 9:09 • According to my understanding these days, GP need a prior (x,y), to build on (x*,y*), however, in my case, I don't have any y, what I have is only x and x*, any further ideas? – cloudscomputes Nov 16 '18 at 8:40 • You can use the month (jan., feb., etc..) as your$x$and the income as your$y$. Afterwards you calculate for missing$x$your corresponding$y\$ with the gp. – Andreas Look Nov 16 '18 at 8:49
• I don't have any income data, I only have person's living cost – cloudscomputes Nov 19 '18 at 3:07