utils.py

from sklearn.model_selection import train_test_split
from sklearn.linear_model import LinearRegression
from sklearn.metrics import mean_squared_error, r2_score
from sklearn.ensemble import RandomForestRegressor
import pandas as pd
from tqdm.auto import tqdm
import streamlit as st
from huggingface_hub import Repository, HfApi, HfFolder
import os
import random
tqdm.pandas()

api = HfApi()
token = os.getenv("token")  # Das Token wird aus den Hugging Face Secrets abgerufen
tokenread = os.getenv("tokenread")
localdir = "SpotifyHitPrediction"

repo = Repository(local_dir=localdir, clone_from="https://huggingface.co/spaces/Add1E/SpotifyHitPrediction", token=token)


def remove_non_utf8_characters(text):
    # Encode using UTF-8, ignore errors
    encoded_string = text.encode('utf-8', 'remove')
    # Decode back to string
    return encoded_string.decode('utf-8')


def predict_popularity(features,trainset):
    predictions = [None] * 2
    predictions[0], predictions[1] = rf_model.predict([features]), model.predict([features])
    old_df = pd.read_csv(f'{localdir}/top50.csv', encoding='utf-8')
    addToCsvAndTrain(trainset, old_df)
    st.write("Regression :")
    st.code(f"MeanSquaredError: {mse}, rSqared: {r2}")
    st.write("Random Forest :")
    st.code(f"MeanSquaredError: {rf_mse}, rSqared: {rf_r2}")
    return predictions


def addToCsvAndTrain(trainset, old_df):
    trainset = [
        [trainset[0], remove_non_utf8_characters(trainset[1]), remove_non_utf8_characters(trainset[2]), trainset[3], trainset[4], trainset[5], trainset[6], trainset[7],
         trainset[8], trainset[9], trainset[10], trainset[11], trainset[12], trainset[13]
         ]
    ]
    neues_df = pd.DataFrame(trainset, columns= data.columns)
    df = pd.concat([old_df, neues_df], ignore_index=True)
    df.to_csv(f'{localdir}/top50.csv', index=False, encoding='utf-8')
    if(random.randint(1, 10) == 7):
        repo.git_add(os.path.abspath(f'{localdir}/top50.csv'))
        repo.git_commit("Add top50.csv")
        repo.git_push()


data = pd.read_csv('top50.csv', encoding='utf-8')
print(data.head())

# Let's also describe the data to get a sense of the distributions
print(data.describe())
# Selecting the features and the target variable
X = data.drop(['Unnamed: 0', 'Track.Name', 'Artist.Name', 'Genre', 'Popularity'], axis=1)
y = data['Popularity']

# Splitting the data into training and testing sets
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)

# Initializing the Linear Regression model
model = LinearRegression()

# Fitting the model
model.fit(X_train, y_train)

# Making predictions
y_pred = model.predict(X_test)

# Calculating the performance metrics
mse = mean_squared_error(y_test, y_pred)
r2 = r2_score(y_test, y_pred)
# Initialize the Random Forest Regressor
rf_model = RandomForestRegressor(n_estimators=100, random_state=42)

# Fitting the model
rf_model.fit(X_train, y_train)

# Making predictions
rf_pred = rf_model.predict(X_test)

# Calculating the performance metrics
rf_mse = mean_squared_error(y_test, rf_pred)
rf_r2 = r2_score(y_test, rf_pred)


# Feature importances
feature_importances = rf_model.feature_importances_

# Create a pandas series with feature importances
importances = pd.Series(feature_importances, index=X.columns)

# Sort the feature importances in descending order
sorted_importances = importances.sort_values(ascending=False)