Please find provided a CSV file containing observed dates for monthly direct debits and standing orders.
Each series of dates has its own series identifier and the date each transaction in the series was observed.
Your task is to predict the next date of the series taking into account the following:
- The simplest solution is preferred, even if it is not the most accurate. Although these two criteria, simplicity and accuracy, need to be balanced.
- No assessment of accuracy needs to be provided, a reasoned argument can be presented when answering the questions below.
- We program in a functional style and the use of RamdaJS is encouraged as we use the library extensively https://ramdajs.com/docs/. Some RamdaJS methods we commonly use are compose, groupBy, map, reduce, aperture and mapAccum amongst others. Sometimes we use a mix of functional and imperative style so do not feel the need to make everything functional, especially if it is at the expense of readability. However, consider immutability a key principle that should be upheld where possible, i.e. avoid reassigning identifiers or mutating objects.
- Results should be produced in CSV or JSON format.
- The exercise should take no more than an hour and reflects the day to day work conducted by the Data Team so hopefully should be fun.
- Describe the method you used to predict the next date and why you chose it?
- What strategy would you use to implement a more accurate date prediction mechanism and what estimate would you give in terms of implementation effort required?
- How would you test the accuracy of date prediction?