I am trying to build a dynamic pricing algorithm on intermittent data (a lot of zeros between non-zero values). I have on average 100 non-zero data points for each product. However, it seems to be quite difficult to predict the right price due to various reasons (non-regular demand, intermittent data, noises etc) so for the first step, I am trying to forecast the demand of each product using MAPA from Nikolaos Kourentzes. I hope to use this demand forecast and price-demand relation to predict the optimal price.

However, in practice these price-demand relationships tend to be non-linear so I was wondering if anyone here could guide me to papers/resources that I can follow to train non-linear model(s) on the price-demand relationship with intermittent demand data. Ideally, the trained model should be optimizable for price.

[Update 23/03/2019 : Added images/tables for elaborating Dynamic pricing]

The input is a intermittent sales data that looks something like the example shown below for most products. Sales data (Intermittent demand image courtesy to)

Using Dynamic pricing, I would like to model the price-demand relation like the one shown below. Important : Linear demand are not realistic and in my case, the demand for my case non-linear. Linear Demand

(Linear demand image from)

In the end, I hope to create a model that can tell me what the optimal price is given a certain demand.

Optimal price

(Optimal demand image from

  • $\begingroup$ Have you tried RNN (LSTM or GRU) models ? These should be able to capture non-linear relationships in time series. $\endgroup$ – Shamit Verma Mar 22 '19 at 4:21
  • $\begingroup$ It is not really clear what you want. Could you give an example what you data looks like and what you want to achieve? $\endgroup$ – MachineLearner Mar 22 '19 at 9:19
  • $\begingroup$ @ShamitVerma I think any RNN models would be an overkill because I do not have much data. Maybe it would have been smarter for mention it in the question. $\endgroup$ – thulungair Mar 23 '19 at 13:26
  • $\begingroup$ @MachineLearner I will edit my question to include a general example $\endgroup$ – thulungair Mar 23 '19 at 14:20
  • $\begingroup$ How much data is availalbe for training ? $\endgroup$ – Shamit Verma Mar 23 '19 at 14:25

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.