len-sla/AutoML

Checking possibilities of AutoML package from MLJAR

based on the https://github.com/mljar/mljar-supervised

---

If someone is looking for ready to use, easy to install( here on lubuntu) automatised ml package for analysis of tabular data this is really excellent choice.
I am really impressed by way of presenting and storage results by mentioned package.
To check and play input data with Moskow flat prices were used.
Typical example for regresion to evaluate the price metric mae.

mljar.ipynb

Below there is starting screen of leaderboard with reasults generated by package:

---

AutoML Leaderboard

Best model	name	model_type	metric_type	metric_value	train_time
	1_Baseline	Baseline	mae	0.225592	2.68
	2_DecisionTree	Decision Tree	mae	0.186864	21.44
the best	3_Default_Xgboost	Xgboost	mae	0.111216	29.3
	4_Default_NeuralNetwork	Neural Network	mae	0.136117	9.68
	5_Default_RandomForest	Random Forest	mae	0.16243	39.82
	Ensemble	Ensemble	mae	0.111216	1.53

AutoML Performance

AutoML Performance Boxplot

Features Importance

Spearman Correlation of Models