How to do Inventory Forecasting using R

r
forecasting
inventory

#1

I want to forecast the inventory using R. I am having variables Date. product id and number of units

i want to forecast it using R . Any suggestion or realtime example is highly appreciated


#2

Hi there,seems to me like what you need is a classic regression ,since time is not of essence.


#3

what about seasonality and cyclic variations? Regression wont consider that thing


#4

Is this a school or work project?
If it is work then have you tried using the R package called prophet?
https://cran.r-project.org/web/packages/prophet/index.html

If it is school what is the requirements for the class? What methods have you used thus far?


#5

no its a work project


#6

Test for regression assumption before you decide seasonality and cyclic variations. Time series is your other alternative.


#7

yes i am thinking about time series but was looking for any alternative


#8

Since you have different variables available in addition to time and quantity data, you can try ARIMA with external regressor. However, I’d suggest you to go with simple time series modelling using ARIMA, Prophet (facebook’s) and NeuralNet, and look at your forecasting accuracy. If that doesn’t workout, then only use the same techniques with external regressors. If you still see that your accuracy isn’t improving, you can consider tuning the ordering of your time series.

There are more advance techniques available as well such as BSTS (google’s) or LSTM (Neural net), but would suggest you to take a look at them once you’ve tried and tested the basics one.

As you pointed out correctly that seasonality, stationarity and cyclicity will be of essence and thus should be incorporated in the model, regular regression can not be used.

Let me know in case you need links to read more about these techniques.

Peace!


#9

I suggest you tell your company you’re struggling with the requirements and need some help. It’s weird, I take on side projects as a Data Scientist and have always researched ways to help me. If you’re a Data Scientist or something similar and represent your company, why is that you need help with a simple forecasting algorithm with simplified variables? I’ll keep fighting for my chance to one day be a fully employed Data Scientist.


#10

brother thats the dark side of being a data scientist . you dont have much help or resource available within your organisation.


#11

I may be completely wrong but just check this stackexchange post if it helps…

https://stats.stackexchange.com/questions/149191/dealing-with-spikes-in-data


#12

I think what you should do first is to define the problem precisely.

What’s you company’s product strategy and competitive pressures ?
Is the Inventory too much?
Is the demand plan always terrible? have impact on supply chain?
You want to only reduce inventory fare ?
After reduction, it can still satisfy customers’ requirements ?
meet rate orders increased frequency ?
Or your goal is only to reduce the inventory capacity shortages ?
…

You should pick one as an optimization problem under some constraints.

What you want to solve is more important than making technology choices
After definition, Combined with actual business requirements,
you can use any modeling software,even only Excel , to model \ test the algorithm.
Sometimes simple rule or model is better than complex algorithm
especially for some cold-start problems or lack of good data.

Last but not least:
Rome was not built in a day
so be able to do just-in-time development, rapid iterations and run multiple experiments

Some tips:
Many related contests are related with inventory,both in AV and Kaggle:
Practice Problem: Black Friday
Practice Problem: Big Mart Sales III
Practice Problem: Time Series
walmart-recruiting-store-sales-forecasting
grupo-bimbo-inventory-demand