andrea-dagostino

import pandas as pd

df = pd.read_csv('wineQualityReds.csv') # download dataset -> https://www.kaggle.com/datasets/piyushgoyal443/red-wine-dataset

# since the dataset contains the target variable in a range between 3 and 8, we map them from 1 to 5.
quality_mapping = {
    3: 0,
    4: 1,
    5: 2,
    6: 3,

andrea-dagostino

train_accs = []
test_accs = []

cols = [
    'fixed.acidity', 'volatile.acidity', 'citric.acid','residual.sugar', 'chlorides', 'free.sulfur.dioxide',
    'total.sulfur.dioxide', 'density', 'pH', 'sulphates', 'alcohol',
]

# inizializziamo un loop dove cambieremo il valore di max depth, partendo da 1 a 25
for depth in range(1, 25):

andrea-dagostino

import pandas as pd

df = pd.read_csv('wineQualityReds.csv') # download dataset -> https://www.kaggle.com/datasets/piyushgoyal443/red-wine-dataset

# poiché il dataset contiene solo i numeri da 3 a 8, rimappiamoli con i numeri da 1 a 5.
quality_mapping = {
    3: 0,
    4: 1,
    5: 2,
    6: 3,

andrea-dagostino

N = 15  # window size --> possiamo modificare questo parametro per sperimentare
K = 0.70 # split size --> 70% dei dati è in A, 30% in B

SEQS = split_time_series(list(data['Close'].values), N) # crea sequenze di lunghezza N

SPLIT_SEQS = split_sequences(SEQS, K) # divide le sequenze in due

A = [seq[0] for seq in SPLIT_SEQS]
B = [seq[1] for seq in SPLIT_SEQS]

andrea-dagostino

def classify_trend(b, threshold=0.05):
    """
    Classify the trend of a vector
    """
    # compute slope
    slope = np.mean(np.diff(b) / np.diff(np.arange(len(b))))
    # if slope is positive, the trend is upward
    if slope + (slope * threshold) > 0:
        return 1
    # if slope is negative, the trend is downward

andrea-dagostino

# populate G
G = {}
THRESHOLD = 6 # arbitrary value - tweak this to get different results

for i in range(len(S)):
    G[i] = []
    for j in range(len(S)):
        if S[i, j] < THRESHOLD and i != j and (i, j) not in G and (j, i) not in G and j not in G[i]:
            G[i].append(j)

andrea-dagostino

# plot heatmap of S
fig, ax = plt.subplots(figsize=(20, 10))  
sns.heatmap(S, cmap='nipy_spectral_r', square=True, ax=ax)
plt.title("Heatmap of sequence similarities", fontsize=20, fontweight='bold')
plt.xticks(range(len(A)), range(len(A)))
plt.yticks(range(len(A)), range(len(A)))
plt.show()

andrea-dagostino

def compute_correlation(a1, a2):
    """
    Calculate the correlation between two vectors
    """
    return np.corrcoef(a1, a2)[0, 1]

def compute_dynamic_time_warping(a1, a2):
    """
    Compute the dynamic time warping between two sequences
    """

andrea-dagostino

N = 15  # window size --> we can tweak it and test different options
K = 0.70 # split size --> 80% of the data is in A, 20% in B

SEQS = split_time_series(list(data['Close'].values), N) # creates sequences of length N

SPLIT_SEQS = split_sequences(SEQS, K) # splits sequences into individual sequences

A = [seq[0] for seq in SPLIT_SEQS]
B = [seq[1] for seq in SPLIT_SEQS]

andrea-dagostino

def split_sequence(sequence, k):
    """
    Split a sequence in two, where k is the size of the first sequence
    """
    return np.array(sequence[:int(len(sequence) * k)]), np.array(sequence[int(len(sequence) * k):])


def split_sequences(sequences, k=0.80):
    """
    Applies split_sequence on all elements of a list or array