|
import streamlit as st |
|
from streamlit_shap import st_shap |
|
import shap |
|
from datasets import load_dataset |
|
from sklearn.model_selection import train_test_split |
|
import lightgbm as lgb |
|
import numpy as np |
|
import pandas as pd |
|
|
|
|
|
@st.experimental_memo |
|
def load_data(): |
|
dataset = load_dataset("ttd22/house-price", streaming = True) |
|
df = pd.DataFrame.from_dict(dataset["train"]) |
|
df = df.drop('Id', axis=1) |
|
drop_columns = (df.isnull().sum().sort_values(ascending=False).loc[lambda x : x > .90*1460]).index.to_list() |
|
df = df.drop(drop_columns, axis = 'columns', errors = 'ignore') |
|
cols_with_missing_values = df.columns[df.isnull().sum() > 0] |
|
|
|
for col in cols_with_missing_values: |
|
|
|
if df[col].dtype in ['int64', 'float64']: |
|
|
|
median = df[col].median() |
|
df[col].fillna(median, inplace=True) |
|
else: |
|
|
|
mode = df[col].mode()[0] |
|
df[col].fillna(mode, inplace=True) |
|
X, y = df.drop("SalePrice", axis=1), df["SalePrice"] |
|
|
|
cat_features = X.select_dtypes(exclude=np.number).columns.to_list() |
|
cat_idx = [X.columns.get_loc(col) for col in cat_features] |
|
|
|
for col in cat_features: |
|
X[col] = pd.Categorical(X[col]) |
|
return X,y,cat_idx |
|
|
|
@st.experimental_memo |
|
def load_model(X, y, cat_idx): |
|
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42) |
|
params = {'n_estimators': 569, 'num_leaves': 62, 'max_depth': 10, 'learning_rate': 0.010786783375710743, 'colsample_bytree': 0.5065493231651268, 'subsample': 0.7900705177300663, 'lambda_l1': 4.998785478697207, 'lambda_l2': 2.1857959934319657, 'min_child_weight': 11.187719709451862} |
|
model = lgb.LGBMRegressor(**params) |
|
model.fit(X_train, y_train, eval_set=[(X_test, y_test)], categorical_feature=cat_idx, verbose = False) |
|
return model |
|
|
|
|
|
X,y,cat_idx = load_data() |
|
model = load_model(X, y, cat_idx) |
|
|