Have a problem that really I am banging my head against the wall and don't know why is happening. I have a file with columns as strings and I am using Manipulation -> Column -> Convert & Replace -> String to Number, in order to make my string values as numbers -because I want to use neural networks for prediction- and when i use and execute it, there is not parsing at all and i do not know why, it is a normal string.
Would appreciate some help on this, thank you all in advance. :D
WARN String To Number 2:14 Values in 4599 cells could not be parsed, first error: '2015/06' (RowKey: Row0, Position: 0)
WARN String To Number 2:14 Problems occurred, see Console messages.
this is very obvious in the log: 2015/06 is not a number ;-)
You could split up the string into a year and month component and then parse both individually. Or parse it as a date type using "String to date" node.
Actually is not a date, it is a string. but also happnes with another strings such as "Component".
sorry if I misunderstood, but then what's your intention for converting those values to numbers? The "String to Number" merely performs a type conversion (string "1" -> number "1"), it's not intended to convert values such as "alpha" to a number representation.
As you're mentioning neural nets, are you probably looking for an encoding of nominal values, such as one-hot?
Yes Philipp, you are correct. What i want to do is make all the strings on my data to be encoded as numbers, because neural nets only work with numbers. For instance, if i have a column named "Component" with values
then make them
get my point?. If so, what would you suggest to use?
Many thanks in advance.
- import a correspondance list between the strings and your numeric codes and merge it with your table;
- create such a list à la mano using Table creator if you want some sort of interface;
- use Rule Engine to do the above à la mano;
- use One To Many node to go the dummy variable route instead.
Probably there are many more possibilities.