Introduction

ANAI

ANAI is an Automated Machine Learning Python Library that works with tabular data. It is intended to save time when performing data analysis. It will assist you with everything right from the beginning i.e Ingesting data using the inbuilt connectors, preprocessing, feature engineering, model building, model evaluation, model tuning and much more.

Let's Democratize AI

In [1]:
%pip install anai-opensource

Importing Libraries

In [ ]:
import anai
from anai.preprocessing import Preprocessor

Loading Data

In [2]:
df = anai.load(
    df_filepath='/Users/arshanwar/Projects/AutoML/open_source/ANAI/examples/healthcare-dataset-stroke-data.csv')

Loading Data [*]

Data Loaded Successfully [ ✓ ]
In [3]:
df.head()
Out[3]:
id gender age hypertension heart_disease ever_married work_type Residence_type avg_glucose_level bmi smoking_status stroke
0 9046 Male 67.0 0 1 Yes Private Urban 228.69 36.6 formerly smoked 1
1 51676 Female 61.0 0 0 Yes Self-employed Rural 202.21 NaN never smoked 1
2 31112 Male 80.0 0 1 Yes Private Rural 105.92 32.5 never smoked 1
3 60182 Female 49.0 0 0 Yes Private Urban 171.23 34.4 smokes 1
4 1665 Female 79.0 1 0 Yes Self-employed Rural 174.12 24.0 never smoked 1

ANAI Preprocessing Pipeline

In [4]:
prep = Preprocessor(dataset = df, target = 'stroke')

Summaries

In [5]:
summary = prep.summary()
summary.head(10)
Out[5]:
Stats
No. of Cells 61320
No. of Variables 12
No. of Records 5110
Missing Cells 0.3 %
Missing Cells Count 201
Duplicacy 0.00 %
Duplicate Cell Count 0
Anomaly Count 256
In [6]:
column_summary = prep.column_summary()
column_summary.head(24)
Out[6]:
id gender age hypertension heart_disease ever_married work_type Residence_type avg_glucose_level bmi smoking_status stroke
Type Error ID column is not allowed NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN
hide True NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN
Type NaN Categorical Numeric Numeric Numeric Categorical Categorical Categorical Numeric Numeric Categorical Numeric
Missing Value % NaN 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 3.9334637964774952 0.0 0.0
Mean NaN NA as column dtype is Categorical 43.23 0.10 0.05 NA as column dtype is Categorical NA as column dtype is Categorical NA as column dtype is Categorical 106.15 28.89 NA as column dtype is Categorical 0.05
Mode NaN NA as column dtype is Categorical 78.00 0.00 0.00 NA as column dtype is Categorical NA as column dtype is Categorical NA as column dtype is Categorical 93.88 28.70 NA as column dtype is Categorical 0.00
Maximum value NaN NA as column dtype is Categorical 82.00 1.00 1.00 NA as column dtype is Categorical NA as column dtype is Categorical NA as column dtype is Categorical 271.74 97.60 NA as column dtype is Categorical 1.00
Median NaN NA as column dtype is Categorical 45.00 0.00 0.00 NA as column dtype is Categorical NA as column dtype is Categorical NA as column dtype is Categorical 91.88 28.10 NA as column dtype is Categorical 0.00
Minimum value NaN NA as column dtype is Categorical 0.08 0.00 0.00 NA as column dtype is Categorical NA as column dtype is Categorical NA as column dtype is Categorical 55.12 10.30 NA as column dtype is Categorical 0.00
Standard Deviation NaN NA as column dtype is Categorical 22.61 0.30 0.23 NA as column dtype is Categorical NA as column dtype is Categorical NA as column dtype is Categorical 45.28 7.85 NA as column dtype is Categorical 0.22
99% Quartile NaN NA as column dtype is Categorical 82.00 1.00 1.00 NA as column dtype is Categorical NA as column dtype is Categorical NA as column dtype is Categorical 240.71 53.40 NA as column dtype is Categorical 1.00
90% Quartile NaN NA as column dtype is Categorical 75.00 0.00 0.00 NA as column dtype is Categorical NA as column dtype is Categorical NA as column dtype is Categorical 192.18 38.90 NA as column dtype is Categorical 0.00
66% Quartile NaN NA as column dtype is Categorical 55.00 0.00 0.00 NA as column dtype is Categorical NA as column dtype is Categorical NA as column dtype is Categorical 104.08 31.00 NA as column dtype is Categorical 0.00
33% Quartile NaN NA as column dtype is Categorical 32.00 0.00 0.00 NA as column dtype is Categorical NA as column dtype is Categorical NA as column dtype is Categorical 81.78 25.10 NA as column dtype is Categorical 0.00
10% Quartile NaN NA as column dtype is Categorical 11.00 0.00 0.00 NA as column dtype is Categorical NA as column dtype is Categorical NA as column dtype is Categorical 65.79 19.70 NA as column dtype is Categorical 0.00
1% Quartile NaN NA as column dtype is Categorical 1.08 0.00 0.00 NA as column dtype is Categorical NA as column dtype is Categorical NA as column dtype is Categorical 56.33 15.10 NA as column dtype is Categorical 0.00
Variance NaN NA as column dtype is Categorical 511.33 0.09 0.05 NA as column dtype is Categorical NA as column dtype is Categorical NA as column dtype is Categorical 2050.60 61.69 NA as column dtype is Categorical 0.05
Monotonic NaN NA as column dtype is Categorical 0.00 0.00 0.00 NA as column dtype is Categorical NA as column dtype is Categorical NA as column dtype is Categorical 0.00 0.00 NA as column dtype is Categorical 0.00
Mean Absolute Deviation NaN NA as column dtype is Categorical 19.12 0.18 0.10 NA as column dtype is Categorical NA as column dtype is Categorical NA as column dtype is Categorical 33.06 5.98 NA as column dtype is Categorical 0.09
No. of Unique Values NaN 3 104 2 2 2 5 2 3979 419 4 2
No. of Negative Values NaN NA as column dtype is Categorical 0 0 0 NA as column dtype is Categorical NA as column dtype is Categorical NA as column dtype is Categorical 0 0 NA as column dtype is Categorical 0
Percentage Infinite Values NaN NA as column dtype is Categorical 0.0 0.0 0.0 NA as column dtype is Categorical NA as column dtype is Categorical NA as column dtype is Categorical 0.0 0.0 NA as column dtype is Categorical 0.0
Skewness NaN NA as column dtype is Categorical -0.14 2.72 3.95 NA as column dtype is Categorical NA as column dtype is Categorical NA as column dtype is Categorical 1.57 1.06 NA as column dtype is Categorical 4.19
Shapiro_W NaN NA as column dtype is Categorical 0.97 0.34 0.24 NA as column dtype is Categorical NA as column dtype is Categorical NA as column dtype is Categorical 0.81 nan NA as column dtype is Categorical 0.22

As we can see there are 0.3% missing cells. We will impute them now.¶

Imputing Missing Values

In [7]:
df1 = prep.impute(method = 'mean')
In [8]:
df1.isna().sum()
Out[8]:
id                   0
gender               0
age                  0
hypertension         0
heart_disease        0
ever_married         0
work_type            0
Residence_type       0
avg_glucose_level    0
bmi                  0
smoking_status       0
stroke               0
dtype: int64

Preparing Data

In [9]:
features, labels = prep.encode(split = True)
In [10]:
features.head(4)
Out[10]:
id gender age hypertension heart_disease ever_married work_type Residence_type avg_glucose_level bmi smoking_status
0 9046 0.048728 67.0 0 1 0.048728 0.048728 0.048728 228.69 36.6 0.048728
1 51676 0.048728 61.0 0 0 0.524364 0.048728 0.048728 202.21 NaN 0.048728
2 31112 0.524364 80.0 0 1 0.682909 0.524364 0.524364 105.92 32.5 0.524364
3 60182 0.524364 49.0 0 0 0.762182 0.682909 0.524364 171.23 34.4 0.048728

#¶

In [11]:
X_train, X_val, y_train, y_val, scaler = prep.prepare(features, labels, test_size = 0.2, random_state = 42, smote = False, k_neighbors = 3)
In [12]:
X_train.shape, X_val.shape, y_val.shape, y_train.shape
Out[12]:
((4088, 11), (1022, 11), (1022,), (4088,))

ANAI's AutoML Pipeline

In [14]:
ai = anai.run(filepath='/Users/arshanwar/Projects/AutoML/open_source/ANAI/examples/healthcare-dataset-stroke-data.csv',
              target='stroke', predictor=['rfc', 'cat', 'xgb', 'lgbm', 'ext'])

ANAITaskWarning: Task is getting detected automatically. To suppress this behaviour, set suppress_task_detection=True and specify task with task argument 
Task: Classification 


░█████╗░███╗░░██╗░█████╗░██╗
██╔══██╗████╗░██║██╔══██╗██║
███████║██╔██╗██║███████║██║
██╔══██║██║╚████║██╔══██║██║
██║░░██║██║░╚███║██║░░██║██║
╚═╝░░╚═╝╚═╝░░╚══╝╚═╝░░╚═╝╚═╝
 

Started ANAI [ ✓ ]

Preprocessing Started [*]

Imputing Missing Values by mean [*]

Imputing Done [ ✓ ]

Preprocessing Done [ ✓ ]

Training ANAI [*]

Ensembling on top 5 models

Training Done [ ✓ ]

Results Below
Name Accuracy Cross Validated Accuracy
0 Random Forest Classifier 99.412916 99.363824
1 Stacking Ensembler 99.510000 99.340000
2 Extra Trees Classifier 99.412916 99.314864
3 Max Voting Ensembler 99.410000 99.310000
4 CatBoost Classifier 99.412916 99.265904
5 XGBoost Classifier 99.412916 99.241574
6 LightGBM Classifier 99.412916 99.143715 

Completed ANAI Run [ ✓ ]

Saved Best Model to anai_info/best/classifier/models/Random_Forest_Classifier_1655501647.pkl and its scaler to anai_info/best/classifier/scalers/Random_Forest_Classifier_Scaler_1655501647.pkl 

Time Elapsed :  144.39 seconds

Explainable ANAI

In [17]:
ai.explain('perm')

Explaining Best ANAI model [*]


Explaining ANAI Done [ ✓ ]

Leaderboard

In [16]:
ai.result()
Out[16]:
Name Accuracy Cross Validated Accuracy Model
0 Random Forest Classifier 99.412916 99.363824 (DecisionTreeClassifier(max_features='auto', r...
1 Stacking Ensembler 99.510000 99.340000 StackingClassifier(cv=10,\n ...
2 Extra Trees Classifier 99.412916 99.314864 (ExtraTreeClassifier(random_state=840703915), ...
3 Max Voting Ensembler 99.410000 99.310000 VotingClassifier(estimators=[('Random Forest C...
4 CatBoost Classifier 99.412916 99.265904 <catboost.core.CatBoostClassifier object at 0x...
5 XGBoost Classifier 99.412916 99.241574 XGBClassifier(base_score=0.5, booster='gbtree'...
6 LightGBM Classifier 99.412916 99.143715 LGBMClassifier()

Final Notes

The purpose of this Notebook was to give overview of ANAI and show how easy it is to use a ANAI.