mirror of
https://git.roussel.pro/telecom-paris/GIN206.git
synced 2026-02-09 02:30:17 +01:00
120 lines
3.9 KiB
Python
120 lines
3.9 KiB
Python
import datetime
|
|
from analyze import hourly_rate_of_change
|
|
|
|
def sample_every_kth_point(df, k):
|
|
"""
|
|
Sample every k-th point from a DataFrame.
|
|
|
|
Parameters:
|
|
- df: pandas DataFrame
|
|
The DataFrame from which to sample the points.
|
|
- k: int
|
|
The interval between sampled points.
|
|
|
|
Returns:
|
|
- sampled_df: pandas DataFrame
|
|
The DataFrame containing the sampled points.
|
|
Raises:
|
|
- ValueError: If k is not a positive integer or if k exceeds the number of rows in the DataFrame.
|
|
"""
|
|
|
|
# Validate the input to ensure k is positive and does not exceed the DataFrame length
|
|
if k <= 0:
|
|
raise ValueError("k must be a positive integer.")
|
|
if k > len(df):
|
|
raise ValueError("k is greater than the number of rows in the DataFrame.")
|
|
|
|
# Sample every k-th point
|
|
sampled_df = df.iloc[::k]
|
|
return sampled_df
|
|
|
|
def optimal_sample(df, threshold_dT=0.5):
|
|
"""
|
|
Returns a subset of the input DataFrame `df` containing rows that have a significant change in value.
|
|
|
|
Parameters:
|
|
df (pandas.DataFrame): The input DataFrame.
|
|
threshold_dT (float, optional): The threshold value for the change in value. Defaults to 0.5.
|
|
|
|
Returns:
|
|
pandas.DataFrame: A subset of the input DataFrame `df` containing rows with significant changes in value.
|
|
"""
|
|
|
|
t0 = df["time"].iloc[0]
|
|
indices = [0]
|
|
times = [t0]
|
|
for i in range(1, len(df)):
|
|
dT = abs(df["value"].iloc[i] - df["value"].iloc[indices[-1]])
|
|
if dT > threshold_dT:
|
|
times.append(i)
|
|
indices.append(i)
|
|
return df.iloc[indices]
|
|
|
|
def sample_reglin(df, max_dT=0.5, max_poll_interval=2 * 3600):
|
|
"""
|
|
Returns a subset of the input DataFrame `df` by sampling points based on a linear regression algorithm.
|
|
|
|
Parameters:
|
|
- df (pandas.DataFrame): The input DataFrame containing the time series data.
|
|
- max_dT (float): The value difference that should be considered significant enough to add a new value.
|
|
Defaults to 0.5.
|
|
- max_poll_interval (int): The maximum time interval allowed between the first and last point in the subset.
|
|
Defaults to 2 hours (2 * 3600 seconds).
|
|
|
|
Returns:
|
|
- pandas.DataFrame: A subset of the input DataFrame `df` containing the sampled points.
|
|
|
|
Raises:
|
|
- ValueError: If there is no point before the specified date.
|
|
|
|
"""
|
|
indices = []
|
|
|
|
def get_first_point_after(date):
|
|
if df[df['time'] > date].empty:
|
|
raise ValueError("No point before the date")
|
|
return df[df['time'] > date].iloc[0]
|
|
|
|
# Get first two points
|
|
t0 = df["time"].iloc[0]
|
|
t1 = df["time"].iloc[1]
|
|
|
|
while True:
|
|
v0 = df[df["time"] == t0]["value"].values[0]
|
|
v1 = df[df["time"] == t1]["value"].values[0]
|
|
|
|
# Calculate the slope
|
|
s = abs((v1 - v0) / (t1 - t0).total_seconds())
|
|
|
|
# Add max_dT/s to t1
|
|
new_t = t1 + datetime.timedelta(seconds=min(max_dT / s, max_poll_interval))
|
|
|
|
try:
|
|
new_t = get_first_point_after(new_t)["time"]
|
|
indices.append(df[df["time"] == new_t].index[0])
|
|
t0 = t1
|
|
t1 = new_t
|
|
except ValueError:
|
|
break
|
|
|
|
return df.loc[indices]
|
|
|
|
def sample_avg_rate_of_change(df, poll_rate):
|
|
"""
|
|
Calculate the sample average rate of change for a given DataFrame.
|
|
|
|
Args:
|
|
df (pandas.DataFrame): The DataFrame containing the data.
|
|
poll_rate (pandas.Series): The Series containing the poll rates for each hour.
|
|
|
|
Returns:
|
|
pandas.DataFrame: The subset of the DataFrame with the indices where the rate of change exceeds the poll rate.
|
|
|
|
"""
|
|
indices = [0]
|
|
for i in range(len(df)):
|
|
current_hour = df["time"].iloc[i].hour
|
|
if df["time"].iloc[i] - df["time"].iloc[indices[-1]] > datetime.timedelta(seconds=poll_rate.iloc[current_hour]):
|
|
indices.append(i)
|
|
return df.iloc[indices]
|