I want to do my first beginner-level project regarding Data Science. I picked my data set which is described below. I want to make a Python script which will make some predictions on our data set using training and test data. That could be for example a prediction which countries will see rise in number of customers and which ones will see a decline, but that's just the first idea that popped in my head. I want to implement a few Machine Learning methods to do the same prediction on our data set and compare their results. Finally comes my question - which Machine Learning methods would you suggest for a beginner Data Science project with this data set which is described below? Thank you in advance for any suggestions. I don't want to pick something very hard as the beginner and I also don't want to pick a method that just doesn't "fit" my data set (because it's rather simple).

EDIT: To specify this project will concern business predictions based probably on time-series. I'm looking for recommendations for which ML methods are easy to implement use and understand for the beginner. Project will be written in Python.

Data Set Description:

Number of Attributes : 8

Invoice : Integer

StockCode : String

Description: String

Quantity : Integer

InvoiceDate : Timestamp

Price : Decimal

CustomerID : Integer

Country : String

Number of records: 541910

Date ranges: 01/12/2010 08:26 - 09/12/2011 12:50

Price ranges: -11062.06 - 38970.00

Number of distinct Countries: 38

  • $\begingroup$ Welcome to DS SE! I think your approach is somewhat backwards. You first want to identify a problem, then identify solutions. Otherwise, this post can be interpreted as "I have this data, what can I do with it". This can prompt nearly infinite answers! I encourage you to post again after deciding on a specific challenge to tackle (e.g. "given xyz inputs I want to predict the next day's [output]". Others can then suggest ML approaches to a particular problem, but they cannot figure out what your problem should be! You did a great job describing your data though. $\endgroup$ – Benji Albert Nov 21 '19 at 20:57
  • $\begingroup$ Additionally, I agree with @Erwan that this dataset is not trivial. Try to start with data that is purely numeric and has a clear "input/output" schema. You might find this post interesting: towardsdatascience.com/… A Bayesian approach might offer an intuitive and interpretable model to your problem. Good luck! $\endgroup$ – Benji Albert Nov 21 '19 at 21:08

That doesn't look at all like a simple dataset for a beginner's project:

  • First if you want to predict future trends you need to look at time series, which is not the easiest type of method. Additionally our data covers only 2 years, it might be too short for good time predictions.
  • Your dataset is made of various types of attributes so it will probably require some serious preprocessing work (e.g. description string) in order to make the data usable for a particular task.
  • Start with plotting your data: volume of sales by country, number of customer by country, most common products, this kind of thing.
|improve this answer|||||

Not the answer you're looking for? Browse other questions tagged or ask your own question.