import pandas as pd
import numpy as np
import matplotlib.pyplot as plt

col_names = ['T(C)', 'Td(C)', 'HR%', 'ff(kmh)', 'Gust(kmh)', 'P_0(HPa)', 'P_sea(HPa)']

data = pd.read_csv(
    "data_meteo.txt",
    sep=r'\s+',
    skiprows=1,
    usecols=[2, 3, 4, 5, 6, 7, 8],
    names=col_names,
    engine='python'
)

print("загружено записей (строк):", len(data))

data = data.apply(pd.to_numeric, errors='coerce')
n_before = len(data)
data = data.dropna(subset=col_names)
n_after = len(data)

print(f"после приведения к числам и dropna: {n_after} строк (удалено {n_before - n_after})")

def correlation(vec1, vec2, center=False):
    v1 = np.asarray(vec1, dtype=float)
    v2 = np.asarray(vec2, dtype=float)

    if center:
        v1 = v1 - np.mean(v1)
        v2 = v2 - np.mean(v2)

    dot_product = np.sum(v1 * v2)
    norm1 = np.sqrt(np.sum(v1 ** 2))
    norm2 = np.sqrt(np.sum(v2 ** 2))

    if norm1 == 0 or norm2 == 0:
        return np.nan

    return dot_product / (norm1 * norm2)

n = len(col_names)

raw_matrix = np.zeros((n, n))
center_matrix = np.zeros((n, n))

for i in range(n):
    for j in range(n):
        r_raw = correlation(data.iloc[:, i], data.iloc[:, j], center=False)
        r_center = correlation(data.iloc[:, i], data.iloc[:, j], center=True)

        raw_matrix[i, j] = np.round(r_raw, 3) if not np.isnan(r_raw) else np.nan
        center_matrix[i, j] = np.round(r_center, 3) if not np.isnan(r_center) else np.nan

df_raw = pd.DataFrame(raw_matrix, index=col_names, columns=col_names)
df_center = pd.DataFrame(center_matrix, index=col_names, columns=col_names)

print("\nкорреляция (raw):")
print(df_raw.to_string())

print("\nкорреляция (centered):")
print(df_center.to_string())

print("\nразница (centered - raw):")
print((df_center - df_raw).round(3).to_string())

plt.figure(figsize=(14, 8))

for col in col_names:
    centered_values = data[col] - data[col].mean()
    plt.plot(centered_values.values, label=col, marker='.', linewidth=1, markersize=4)

plt.axhline(0, linestyle='--')
plt.title("центрированные метеоданные")
plt.xlabel("номер измерения")
plt.ylabel("отклонение от среднего")
plt.legend()
plt.grid()
plt.tight_layout()

plt.savefig("meteo_analysis.png", dpi=150)
plt.show()