AutoML: “automatic” data processing. Machine training

AI & Machine Learning — incorrectly specified categories?

20 USD

Client's review of cooperation with freelancer

AutoML: “automatic” data processing. Machine training

20 USD

Interesting executive. He completed the project very quickly, used the required material. I have learned a lot of new about this type of work. He answered a lot of questions. Only positive emotions from work. I would recommend and will definitely do it again.

5.0

Alex Alex

Reciprocal Feedback

16 February 2021

Freelancer's review of cooperation with Alex Alex

AutoML: “automatic” data processing. Machine training

20 USD

Clean TZ, timely payment

5.0

Profile deleted

Reciprocal Feedback

16 February 2021

Project translated automatically. Log in or register, to view the original It is necessary to process the network and create automatic algorithms for learning and processing data sets.Implementation of a metaheuristic algorithm for automated selection/configuration of machine learning algorithms Implement a metaheuristic algorithm (local search, genetic algorithms...) that searches for the best machine learning technique (and best hyperparameters) for a particular classification/regression data set.Basic points to be implemented in the work: By taking the data set and cleansing it, we prepare (https://www.kaggle.com/jsphyg/weather-dataset-rattle-package - use the most updated version of the data set 2020) We create algorithms for the possibility of filling the missing data, we “upgrade” the base We prepare the required algorithms for automatic learning (At least five available machine learning algorithms.In this point, we realize #K-nearest neighbors #Random Forest # Gradient boosting # Decision tree # Linear SVM as well as additional Multinomial Logistic Regression algorithms automatically search for the best algorithm and the best hyperparameters (Most important hyperparameters that should be tuned for each of these algorithms.You can specify for each hyperparameter a reasonable range of possible values.The goal is to find a solution (the best algorithm/hyperparameters) in the search space that optimizes an evaluation score (e.g., classification accuracy or RMSE).The same approach is already applied from Python’s auto-sklearn libraries, TPOT (Compare you approach with two state of the art AutoML systems (e.g. Auto-sklearn and TPOT.We analyze these indicators We compare our indicators and indicators auto-sklearn, TPOT (we make a conclusion who is "best") For comparison, you can use works that have been created in "manual mode" : https://www.kaggle.com/rodrigofrb/feature-selections-and-classification-models Thro the code I ask you to make notes (what is done) + The more "tables" and "graphics", the better ... for example, you can see in the manual mode ConfusionMatrix which would be the "best Features" to analyze the date, and then see what to choose "automatic".Or immediately set the automatic and display what it is choosing and why (a table or data chart).that is.As a result, it will need to create a "presentation" and the code in the presentation is not required, but graphics are required to prove one or another action.