atml
atml tries to address the following problem: Given a new data set, is there any toolkit that could help me quickly and conveniently come up with a good baseline model?
try it out
- Install from PyPi
pip install atml- Load the data set or use your own one
import pandas as pd
df = pd.read_csv("./test/data/binary_data.csv")
X = df.drop('Survived', axis=1)
label_columns = ['Survived']
y = df[label_columns]- Instantiate and run with AtmlController with default
from atml import AtmlController
atml_c = AtmlController(with_default=True)
atml_c.run(X, y)- Register a new Model and Hyper parameter space for tuning
from atml import AtmlOrchestrator
atml_o = AtmlOrchestrator(with_default=True)
from sklearn.svm import SVC
sp = [
{"property": "kernel", "type": "choice", "value": ["linear", "rbf"]},
{"property": "gamma", "type": "choice", "value": ["scale", "auto"]}
]
atml_o.auto_learning_socket.register(SVC(), sp)
atml_o.run(X, y)- Get the best model
atml_c.get_best_model_details()['MODEL']
