Performing Advanced Feature Engineering in Keras Ulasan

Nikita K. · Diulas hampir 5 tahun lalu

archana g. · Diulas hampir 5 tahun lalu

Yuan L. · Diulas hampir 5 tahun lalu

George N. · Diulas hampir 5 tahun lalu

Maciej M. · Diulas hampir 5 tahun lalu

Julio T. · Diulas hampir 5 tahun lalu

JASPAL V. · Diulas hampir 5 tahun lalu

Pvrmallikharjuna rao P. · Diulas hampir 5 tahun lalu

Juliana B. · Diulas hampir 5 tahun lalu

Francesco N. · Diulas hampir 5 tahun lalu

Nikita K. · Diulas hampir 5 tahun lalu

What happened to the processed timestamp feature?

Marcus W. · Diulas hampir 5 tahun lalu

Manfred S. · Diulas hampir 5 tahun lalu

Shubham K. · Diulas hampir 5 tahun lalu

Murilo A. · Diulas hampir 5 tahun lalu

Chan S. · Diulas hampir 5 tahun lalu

kazokada k. · Diulas hampir 5 tahun lalu

Oskar F. · Diulas hampir 5 tahun lalu

Iago A. · Diulas hampir 5 tahun lalu

John K. · Diulas hampir 5 tahun lalu

Eric T. · Diulas hampir 5 tahun lalu

Good!

Alex C. · Diulas hampir 5 tahun lalu

1. The provided csv does not match the expected column order in the notebook. It ends up reading passenger_count as if it was pickup_datetime, giving dates such as "1" and "5" -- which are not parseable as dates 2. The lab asks the student to parse pickup datetime to add it as a feature column, but ultimately that is not possible due to the input data -- the parsed picks the wrong column as pickup_datetime, and the csv does not have a pickup_datetime column which could be parsed as a result, the student will waste time trying to parse invalid columns and probably won't be able to get to the point where they would get to use such columns for training. The lab solution currently present in git (2021-04-06) gets around this problem by not using the pickup_datetime column at all -- it just defines functions to parse datetime, then drops pickup_datetime from the input. The proposed model ends up using a temporal feature completely by accident: it uses passenger_count as a feature -- but on the csv, the column taken as passenger_count is actually hourofday, already present in the dataset. As a result, the model inputs are incorrectly named, and predicting with passenger_count=5 is actually taking into account the predicted price for a ride at 5am

Rodrigo K. · Diulas hampir 5 tahun lalu

Samuel Angel R. · Diulas hampir 5 tahun lalu

Manu M. · Diulas hampir 5 tahun lalu