def process_data(data):
(dep, model_data) = data
print("Computing model for " + dep)
gm = GenerativeModel(dep, model_data)
gm.sample()
result = summarize_inference_data(gm.inference_data)
return (dep, result)
def main():
# Read in raw Daily data
log.info("Downloading new data...")
new_cases, new_tests, tx_data = get_tx_covid_data()
# Get start and end dates
LAST_DAY = max(new_cases.columns[-1], new_tests.columns[-1])
log.info(f"County data exists through {LAST_DAY.date()}")
log.info(
f"Covid tracking project data exists through {tx_data.index[-1].date()}"
)
if LAST_DAY.date() != tx_data.index[-1].date():
LAST_DAY = min(LAST_DAY, tx_data.index[-1])
log.warning(f"Date mismatch, using {LAST_DAY.date()}")
# Firestore results keyed by time of model run
fs_doc = datetime.now().strftime("%Y-%m-%d")
log.info("Running regional models")
for region, counties in METROS.items():
log.info(region)
df = pd.concat(
[new_cases.loc[counties].sum(), new_tests.loc[counties].sum()],
axis=1)
df.columns = ["positive", "total"]
start_smooth = LAST_DAY - pd.Timedelta(days=7)
end_smooth = LAST_DAY - pd.Timedelta(days=1)
# Fill-in missing test totals from state-wide data
if np.isnan(df.loc[LAST_DAY, "total"]):
df.loc[LAST_DAY, "total"] = (
tx_data.loc[LAST_DAY, "total"] *
df.loc[start_smooth:end_smooth, "total"].sum() /
tx_data.loc[start_smooth:end_smooth, "total"].sum())
# If any mistaken values result in negatives, zero out so model ignores
df["positive"] = np.where(
df["positive"].values > 0,
df["positive"].values,
np.zeros_like(df["positive"].values),
)
gm = GenerativeModel(region, df.loc[:LAST_DAY])
gm.sample()
to_firestore(fs_doc,
{region: summarize_inference_data(gm.inference_data)})
to_firestore(fs_doc, {"updated": time.time()})
def task_render_region_result(country: str, region: str,
run_date: pd.Timestamp):
""" Render a CSV with summary output for a given region """
az.rcParams["data.load"] = "eager"
with tempfile.NamedTemporaryFile() as fp:
s3.Bucket(S3_BUCKET).download_file(
get_inference_data_key(run_date, region, country=country), fp.name)
fp.seek(0)
inference_data = az.from_netcdf(fp.name)
summary = summarize_inference_data(inference_data)
key = get_state_output_key(run_date, region, country=country)
with fs.open(f"{S3_BUCKET}/{key}", "w") as file:
summary.to_csv(file)
def train(region):
processor = data_processor.factory(region)
cache = Cache(BUCKET, f"cache/{region}", f"/tmp/rt-{region}")
timezone = pytz.timezone(TIMEZONES[region])
if FORCE_USE_CACHE or cache.is_fresh():
cache.download()
model, result = itemgetter("model", "result")(cache.get())
updated_at = cache.modified_at().astimezone(timezone)
summary = _build_summary(result, updated_at)
return model, result, summary
df = processor.process_data(START_DATE)
model = GenerativeModel(region, df.loc[region])
model.sample()
result = summarize_inference_data(model.inference_data)
if SET_CACHE:
cache.set({"model": model, "data": df, "result": result})
updated_at = datetime.now().astimezone(timezone)
summary = _build_summary(result, updated_at)
return model, result, summary
""" Modelo para las 15 provincias con más pruebas procesadas"""
""" Distrito Nacional"""
df = get_and_process_covidtracking_data(
run_date=pd.Timestamp.today())
regiones= ["Distrito Nacional","Santo Domingo","Santiago","La Vega", "La Altagracia","Duarte",
"San Cristóbal","La Romana","Puerto Plata","San Pedro de Macorís","San Juan",
"Azua","Peravia","Espaillat","Sánchez Ramírez"]
for i in regiones:
region=i
model_data = df.loc[region]
gm = GenerativeModel(region, model_data)
gm.sample()
result = summarize_inference_data(gm.inference_data)
fig, ax = plt.subplots(figsize=(12, 8))
result.infections.plot(c="C2", label="Expected primary infections")
result.test_adjusted_positive.plot(c="C0", label="Expected positive tests if tests were constant")
result.test_adjusted_positive_raw.plot(c="C1", alpha=.5, label="Expected positive tests", style="--")
gm.observed.positive.plot(c="C7", alpha=.7, label="Reported positive tests")
fig.set_facecolor("w")
ax.legend();
ax.set(title=f"rt.live model inference for {region}", ylabel="number of cases")
sns.despine();
fig, ax = plt.subplots(figsize=(12, 8))
ax.set(title=f"Tasa de Reproducción Efectiva Para {region}", ylabel="$R_e(t)$")
samples = gm.trace["r_t"]