tm_to_wgs84.py

# %%
import pandas as pd
import seaborn as sns

from sklearn.model_selection import train_test_split
from sklearn.preprocessing import StandardScaler
from sklearn.linear_model import LinearRegression
from sklearn.metrics import r2_score

# %%
df = pd.read_csv('./vertex_latlng.csv')
df.head(5)

# %%
df_train_full, df_test = train_test_split(df[['x', 'y', 'lat', 'lng']], test_size=0.2)
df_train, df_valid = train_test_split(df_train_full, test_size=0.3)

# %%
scaler = StandardScaler()
scaler.fit(df_train[['x', 'y']])
x_train = scaler.transform(df_train[['x', 'y']])

# %%
def fit_model(y_true):
    lr = LinearRegression()
    lr.fit(x_train, y_true)
    return lr

def predict(lr, x_feat):
    x_input = scaler.transform(x_feat)
    return lr.predict(x_input)

# %%
lr_lng = fit_model(df_train['lng'])
lr_lat = fit_model(df_train['lat'])
lng_pred = predict(lr_lng, df_valid[['x', 'y']])
lat_pred = predict(lr_lat, df_valid[['x', 'y']])

r2_score(df_valid['lng'], lng_pred), r2_score(df_valid['lat'], lat_pred)


# %%
lng_pred = predict(lr_lng, df_test[['x', 'y']])
lat_pred = predict(lr_lat, df_test[['x', 'y']])

df_result = pd.DataFrame({
    'x': df_test['x'],
    'y': df_test['y'],
    'lat': lat_pred,
    'lng': lng_pred
})
df_result.to_csv('transformed.csv', index=False)