Update exercice_fr.Rmd

module3/exo2/exercice_fr.Rmd

----
-title: "Analyse de l'incidence de la varicelle"
-author: "Votre nom"
-date: "`r Sys.Date()`"
-output:
-  html_document:
-    toc: true
----
-
-```{r setup, include=FALSE}
-knitr::opts_chunk$set(echo = TRUE)
-Préparation des données
-
-Les données de la varicelle proviennent du Réseau Sentinelles.
-Pour assurer la reproductibilité, nous conservons l’URL d’origine mais utilisons une copie locale des données.
-# URL d'origine (traçabilité)
-
-data_url <- "https://www.sentiweb.fr/datasets/all/inc-25-PAY.json"
-
-# Fichier local
-
-local_file <- "incidence-VARI-3.csv"
-
-# Télécharger uniquement si le fichier local n'existe pas
-
-if (!file.exists(local_file)) {
-download.file(data_url, destfile = local_file, mode = "wb")
-}
-
-# Lecture du fichier CSV local (la première ligne est un commentaire)
-
-data <- read.csv(local_file, skip = 1)
-head(data)
-tail(data)
-Conversion des numéros de semaine
-
-Les semaines sont au format ISO-8601. Nous les convertissons en dates
-(correspondant au lundi de chaque semaine).
-library(parsedate)
-
-convert_week <- function(w) {
-ws <- paste(w)
-iso <- paste0(substr(ws, 1, 4), "-W", substr(ws, 5, 6))
-as.character(parse_iso_8601(iso))
-}
-
-data$date <- as.Date(convert_week(data$week))
-data <- data[order(data$date),]
-Vérification : les dates doivent être espacées de 7 jours.
-
-Inspection des données
-
-On utilise le taux d’incidence pour 100 000 habitants (inc100).
-plot(data$date, data$inc100, type = "l",
-xlab = "Date", ylab = "Incidence hebdomadaire (pour 100 000)")
-Incidence annuelle (année épidémique)
-
-Conformément à l’énoncé, chaque période annuelle est définie
-du 1er septembre de l’année N-1 au 1er septembre de l’année N.
-pic_annuel <- function(annee) {
-debut <- paste0(annee - 1, "-09-01")
-fin   <- paste0(annee,     "-09-01")
-semaines <- data$date > debut & data$date <= fin
-sum(data$inc100[semaines], na.rm = TRUE)
-}
-annees <- (as.integer(format(min(data$date), "%Y")) + 1) :
-as.integer(format(max(data$date), "%Y"))
-
-inc_annuelle <- data.frame(
-annee = annees,
-incidence = sapply(annees, pic_annuel)
+import os
+import re
+import csv
+from datetime import date
+import pandas as pd
+
+DATA_PATH = r"C:\Users\nargi\Downloads\inc-25-PAY.csv"
+
+if not os.path.exists(DATA_PATH):
+    raise FileNotFoundError(f"File not found: {DATA_PATH}")
+
+# ------------------------------------------------------------
+# 1) Load Sentinelles CSV robustly (skip broken metadata / find real header)
+# ------------------------------------------------------------
+with open(DATA_PATH, "r", encoding="utf-8", errors="replace") as f:
+    lines = f.read().splitlines()
+
+# Find header line: contains letters + separators and NOT just the metadata blob
+header_idx = None
+for i, ln in enumerate(lines):
+    if not ln.strip():
+        continue
+    if ln.lstrip().startswith("#"):
+        continue
+    low = ln.lower()
+    if ((";" in ln) or ("," in ln) or ("\t" in ln) or ("|" in ln)) and re.search(r"[A-Za-z_]", ln):
+        header_idx = i
+        break
+
+if header_idx is None:
+    raise ValueError("Could not find a header row (CSV table) in the file.")
+
+# detect delimiter from that header line
+header_line = lines[header_idx]
+try:
+    dialect = csv.Sniffer().sniff(header_line, delimiters=[",", ";", "\t", "|"])
+    sep = dialect.delimiter
+except Exception:
+    sep = ";"  # common in FR exports
+
+content = "\n".join(lines[header_idx:])
+df = pd.read_csv(pd.io.common.StringIO(content), sep=sep, engine="python", on_bad_lines="skip")
+
+df = df.loc[:, ~df.columns.astype(str).str.match(r"^Unnamed")]
+if df.empty:
+    raise ValueError("Parsed dataframe is empty. The file may not contain tabular data.")
+
+# ------------------------------------------------------------
+# 2) Identify week column and incidence column
+# ------------------------------------------------------------
+def score_week_col(s: pd.Series) -> float:
+    x = s.dropna().astype(str).str.strip().head(200)
+    if x.empty:
+        return 0.0
+    patterns = [
+        r"^\d{6}$",                 # YYYYWW
+        r"^\d{4}-?W\d{1,2}$",        # YYYY-WW or YYYYWww
+        r"^\d{4}s\d{1,2}$",          # YYYYsWW
+    ]
+    return max(x.str.match(p, case=False).mean() for p in patterns)
+
+week_col = max(df.columns, key=lambda c: score_week_col(df[c]))
+if score_week_col(df[week_col]) < 0.3:
+    raise ValueError(f"Could not detect week column. Columns: {list(df.columns)}")
+
+# incidence column: prefer inc100 like in your R code, else pick best numeric column containing 'inc'
+lower_to_orig = {str(c).strip().lower(): c for c in df.columns}
+inc_col = None
+for k in ["inc100", "inc_100", "incidence100", "incidence_100"]:
+    if k in lower_to_orig:
+        inc_col = lower_to_orig[k]
+        break
+
+def to_num(s: pd.Series) -> pd.Series:
+    return pd.to_numeric(s.astype(str).str.replace(",", ".", regex=False), errors="coerce")
+
+if inc_col is None:
+    cands = []
+    for c in df.columns:
+        if c == week_col:
+            continue
+        if "inc" in str(c).lower():
+            coerced = to_num(df[c])
+            if coerced.notna().sum() > 0:
+                df[c] = coerced
+                cands.append(c)
+    if not cands:
+        # fallback: any numeric column
+        for c in df.columns:
+            if c == week_col:
+                continue
+            coerced = to_num(df[c])
+            if coerced.notna().sum() > 0:
+                df[c] = coerced
+                cands.append(c)
+    if not cands:
+        raise ValueError("Could not find any numeric incidence column.")
+    inc_col = max(cands, key=lambda c: df[c].var(skipna=True))
+
+df[inc_col] = to_num(df[inc_col])
+
+# ------------------------------------------------------------
+# 3) Convert ISO week -> Monday date (like your R parse_iso_8601 approach)
+# ------------------------------------------------------------
+def week_to_monday(w) -> pd.Timestamp:
+    w = str(w).strip()
+
+    m = re.match(r"^(?P<y>\d{4})(?P<w>\d{2})$", w)  # YYYYWW
+    if m:
+        y = int(m.group("y")); ww = int(m.group("w"))
+        return pd.Timestamp(date.fromisocalendar(y, ww, 1))
+
+    m = re.match(r"^(?P<y>\d{4})-?W(?P<w>\d{1,2})$", w, flags=re.IGNORECASE)  # YYYYWww
+    if m:
+        y = int(m.group("y")); ww = int(m.group("w"))
+        return pd.Timestamp(date.fromisocalendar(y, ww, 1))
+
+    m = re.match(r"^(?P<y>\d{4})s(?P<w>\d{1,2})$", w, flags=re.IGNORECASE)  # YYYYsww
+    if m:
+        y = int(m.group("y")); ww = int(m.group("w"))
+        return pd.Timestamp(date.fromisocalendar(y, ww, 1))
+
+    ts = pd.to_datetime(w, errors="coerce")
+    if pd.isna(ts):
+        raise ValueError(f"Unrecognized week format: {w}")
+    return ts
+
+df["date"] = df[week_col].apply(week_to_monday)
+df = df.sort_values("date").reset_index(drop=True)
+
+# ------------------------------------------------------------
+# 4) Define epidemic year EXACTLY as: (Sep 1 N-1, Sep 1 N]
+#    We label it by the END year N (like your R variable 'annee')
+# ------------------------------------------------------------
+def season_end_year(d: pd.Timestamp) -> int:
+    # If date is strictly after Sep 1 of its calendar year => belongs to season ending next year
+    sep1 = pd.Timestamp(f"{d.year}-09-01")
+    return d.year + 1 if d > sep1 else d.year
+
+df["annee"] = df["date"].apply(season_end_year)
+
+# Keep reasonable years (at least one full season)
+valid_years = sorted(df["annee"].unique())
+# ------------------------------------------------------------
+# 5) Compute BOTH metrics per epidemic year:
+#    - cumulative = sum(inc100)
+#    - peak = max(inc100)
+# ------------------------------------------------------------
+annual = (
+    df.groupby("annee", as_index=False)
+      .agg(
+          incidence_sum=(inc_col, "sum"),
+          incidence_peak=(inc_col, "max"),
+          n_weeks=(inc_col, "count"),
+          first_date=("date", "min"),
+          last_date=("date", "max"),
+      )
+      .sort_values("annee")
 )
 
-head(inc_annuelle)
-Résultats demandés
-Année avec l’épidémie la plus forte
-
-inc_annuelle[which.max(inc_annuelle$incidence), ]
-inc_annuelle[which.min(inc_annuelle$incidence), ]
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
+# strongest/weakest by SUM
+strong_sum = annual.loc[annual["incidence_sum"].idxmax(), ["annee", "incidence_sum"]]
+weak_sum   = annual.loc[annual["incidence_sum"].idxmin(), ["annee", "incidence_sum"]]
 
+# strongest/weakest by PEAK
+strong_peak = annual.loc[annual["incidence_peak"].idxmax(), ["annee", "incidence_peak"]]
+weak_peak   = annual.loc[annual["incidence_peak"].idxmin(), ["annee", "incidence_peak"]]
 
+print(f"✅ Loaded: {DATA_PATH}")
+print(f"✅ Separator: {repr(sep)}")
+print(f"✅ Week column: {week_col}")
+print(f"✅ Incidence column: {inc_col}")
+print(f"✅ Date range: {df['date'].min().date()} -> {df['date'].max().date()}")
+print("\n--- Epidemic year definition: (Sep 1 N-1, Sep 1 N] labeled by N ---\n")
 
+print("RESULTS using CUMULATIVE incidence (SUM over the epidemic year):")
+print(f"  Strongest: year {int(strong_sum['annee'])}  | sum = {strong_sum['incidence_sum']:.3f}")
+print(f"  Weakest:   year {int(weak_sum['annee'])}  | sum = {weak_sum['incidence_sum']:.3f}")
 
+print("\nRESULTS using PEAK incidence (MAX weekly over the epidemic year):")
+print(f"  Strongest: year {int(strong_peak['annee'])} | peak = {strong_peak['incidence_peak']:.3f}")
+print(f"  Weakest:   year {int(weak_peak['annee'])} | peak = {weak_peak['incidence_peak']:.3f}")
 
+print("\n--- Sanity check (first 12 epidemic years) ---")
+print(annual.head(12).to_string(index=False))