EuroPython2022/Model-Recommendation

1import gradio as gr2import pandas as pd3import numpy as np4 5from sklearn.preprocessing import LabelEncoder6from sklearn.pipeline import Pipeline7from sklearn.model_selection import GridSearchCV8from sklearn.model_selection import train_test_split9from sklearn.preprocessing import StandardScaler10 11from sklearn.linear_model import LinearRegression12from sklearn.svm import SVR13from sklearn.tree import DecisionTreeRegressor14from sklearn.ensemble import RandomForestRegressor15 16from sklearn.linear_model import LogisticRegression17from sklearn.neighbors import KNeighborsClassifier18from sklearn.svm import SVC19from sklearn.tree import DecisionTreeClassifier20from sklearn.ensemble import RandomForestClassifier21 22def read(file,dep,ord):23    df = pd.read_csv(file.name)24    cat = list()25    dep_type = str(df.dtypes[dep])26    for col in df.columns.values:27        if str(df.dtypes[col]) == 'bool' or str(df.dtypes[col]) == 'object':28            cat.append(col)29    new_df = df.dropna(axis=0)30    if ord == "" and (dep_type == 'bool' or dep_type == 'object'):31        ord = list()32        ord.append(dep)33    elif ord == "":34        ord = list()35    else:36        pass37    if len(ord)!=0:38        le = LabelEncoder()39        new_df[ord] = new_df[ord].apply(lambda col: le.fit_transform(col))40    nom = list(set(cat).difference(set(ord)))41    if len(nom) == 0:42        pass43    else:44        ohe_df = pd.get_dummies(new_df[nom], drop_first=True)45        new_df.drop(columns=nom, axis=1,inplace=True)46        new_df = pd.concat([new_df,ohe_df],axis=1)47    if dep_type == 'bool' or dep_type == 'object':48        text = "classification"49        result = classification(new_df,dep)50    else:51        text = "regression"52        result = regression(new_df,dep)53    return df.head(5),new_df.head(5),result, text, cat, ord, nom54    55def classification(df,dep):56    X = df.drop(dep,axis=1)57    y = df[dep]58 59    X_train, X_test, y_train, y_test = train_test_split(X, y)60 61    scale = StandardScaler()62 63    pipe = Pipeline(steps=[('scale',scale),('classification','pass')])64 65    parameters = [66        {67            'classification':[LogisticRegression()],68        },69        {70            'classification':[RandomForestClassifier()],71        },72        {73            'classification':[DecisionTreeClassifier()],74        },75        {76            'classification':[SVC()],77        },78        {79            'classification':[KNeighborsClassifier(n_neighbors=5)],80        },81    ]82 83    search = GridSearchCV(pipe, param_grid=parameters, n_jobs=-1, scoring='accuracy')84    search.fit(X_train,y_train)85 86    result = pd.DataFrame(search.cv_results_)[['params','rank_test_score','mean_test_score']]87 88    result['mean_test_score']= (result['mean_test_score'])*10089    result = result.astype({'params': str})90 91    result.sort_values('rank_test_score',inplace=True)92    return result93 94def regression(df,dep):95    X = df.drop(dep,axis=1)96    y =df[dep]97 98    X_train, X_test, y_train, y_test = train_test_split(X, y)99 100    scale = StandardScaler()101 102    pipe = Pipeline(steps=[('scale',scale),('regression','pass')])103 104    parameters = [105        {106            'regression':[LinearRegression()]107        },108        {109            'regression':[RandomForestRegressor()],110        },111        {112            'regression':[DecisionTreeRegressor()],113        },114        {115            'regression':[SVR()],116        },117    ]118 119    search = GridSearchCV(pipe, param_grid=parameters, cv=5, n_jobs=-1, scoring='neg_mean_absolute_percentage_error')120    search.fit(X_train,y_train)121 122    result = pd.DataFrame(search.cv_results_)[['params','rank_test_score','mean_test_score']]123    124    result['mean_test_score']= (result['mean_test_score']+1)*100125    result = result.astype({'params': str})126    127    result.sort_values('rank_test_score',inplace=True)128    return result129    130 131with gr.Blocks() as demo:132    gr.Markdown("Model Recommendation App **Upload** file to see the output.")133    with gr.Column():134        with gr.Row():135            file = gr.File(label="Upload File(Comma Separated)")136            dep = gr.Textbox(label="Dependent Variable(Variable as in the file)")137            ord = gr.Textbox(label="Ordinal Variables(Seperate with a comma)")138            submit = gr.Button("Submit")139        text = gr.Text(label="Suitable Algorithm")140        other1 = gr.Text(label="Categorical Variables")141        other2 = gr.Text(label="LabelEncoded Vairables")142        other3 = gr.Text(label="OneHotEncoded Variables")143        with gr.Row():144            org = gr.DataFrame(overflow_row_behaviour="paginate", label="Original Data")145            converted = gr.DataFrame(overflow_row_behaviour="paginate", label="Transformed Data")146        result = gr.DataFrame(label="Result")147    submit.click(fn=read, inputs=[file,dep,ord], outputs=[org,converted,result,text,other1,other2,other3])148demo.launch()149