def weather_pred(city: str, state: str, metric=None):
db = PostgreSQL()
conn = db.connection
cur = conn.cursor()
db.adapters(np.int64, np.float64, np.datetime64)
# If prediciton found in database
# If metric values are desired:
if metric == True:
table = "feelslikec"
else:
table = "feelslikef"
retrieve_pred = f"""
SELECT month, mean
FROM {table}
WHERE "city"='{city}' and "state"='{state}'
"""
cur.execute(retrieve_pred)
result = pd.DataFrame.from_records(
cur.fetchall(), columns=["month", "mean"])
result.set_index("month", inplace=True)
result.index = pd.to_datetime(result.index)
if len(result.index) > 0:
c = pd.Series([city] * len(result.index))
c.index = result.index
s = pd.Series([state] * len(result.index))
s.index = result.index
result = pd.concat([c, s, result], axis=1)
result.columns = ["city", "state", "temp"]
# If prediction not found in database
elif len(result.index) == 0:
retrieve_data = f"""
SELECT date_time, FeelsLikeC
FROM historic_weather
Where "city"='{city}' and "state"='{state}'
"""
cur.execute(retrieve_data)
df = pd.DataFrame.from_records(cur.fetchall())
df.set_index(0, inplace=True)
series = df.resample("MS").mean()
warnings.filterwarnings(
"ignore",
message="After 0.13 initialization must be handled at model creation"
)
result = ExponentialSmoothing(
series.astype(np.int64),
trend="add",
seasonal="add",
seasonal_periods=12
).fit().forecast(24)
c = pd.Series([city] * len(result.index))
c.index = result.index
s = pd.Series([state] * len(result.index))
s.index = result.index
result = pd.concat([c, s, result], axis=1)
result.columns = ["city", "state", "temp"]
result.index = result.index.astype(str)
# Converting for temperature in Fahrenheit if needed
# Conversion Helper Function
def to_fahr(temp: float) -> float:
return ((temp * 9) / 5) + 32
if metric != True:
result["temp"] = result["temp"].apply(to_fahr)
insert_pred = """
INSERT INTO {table}(
month,
city,
state,
mean
) VALUES%s
""".format(table=table)
execute_values(
cur,
insert_pred,
list(result.to_records(index=True))
)
conn.commit()
conn.close()
return result.to_json(indent=2)
async def predict_temperatures(city: str, state: str, metric=False):
"""
**Input**
`city: str` <- city name with any capitalization
`state: str` <- two-letter state abbreviation with any capitalization
`metric: bool` <- *(Optional)* default, False, will output predictions and make
database queries for adjusted temperature in degrees Fahrenheit;
True will do so for adjusted temperature in degrees Celsius.
**Output**
`json object` with predictions for
**monthly temperature** of `city` and up to a total of three cities
adjusted for dew point and wind chill for 24 months
from the present (September 2020)
"""
db = PostgreSQL()
conn = db.connection
cur = conn.cursor()
db.adapters(np.int64, np.float64, np.datetime64)
# Edge Cases
# Saint
if city[0:3] == "St." or city[0:3] == "st.":
city = city.replace(city[0:3], "St")
elif city[0:5] == "Saint" or city[0:5] == "saint":
city = city.replace(city[0:5], "St")
# Fort
elif city[0:3] == "Ft " or city[0:3] == "ft ":
city = city.replace(city[0:3], "Fort ")
elif city[0:3] == "Ft." or city[0:3] == "ft.":
city = city.replace(city[0:3], "Fort")
# multiple caps
elif city[0:2] == 'Mc':
city = city.replace(city, city[:2] + city[2:].capitalize())
# If prediciton found in database
# If metric values are desired:
if metric == True:
table = "feelslikec"
else:
table = "feelslikef"
retrieve_pred = f"""
SELECT month, mean
FROM {table}
WHERE "city"='{city.title()}' and "state"='{state.upper()}'
"""
cur.execute(retrieve_pred)
result = pd.DataFrame.from_records(cur.fetchall(),
columns=["month", "mean"])
result.set_index("month", inplace=True)
result.index = pd.to_datetime(result.index)
result_json = result.to_json(index=2)
if len(result.index) > 0:
c = pd.Series([city] * len(result.index))
c.index = result.index
s = pd.Series([state] * len(result.index))
s.index = result.index
result = pd.concat([c, s, result], axis=1)
result.columns = ["city", "state", "temp"]
# If prediction not found in database
elif len(result.index) == 0:
retrieve_data = f"""
SELECT date_time, FeelsLikeC
FROM historic_weather
Where "city"='{city.title()}' and "state"='{state.upper()}'
"""
cur.execute(retrieve_data)
df = pd.DataFrame.from_records(cur.fetchall())
if len(df.index) == 0:
result_json = f"{city}, {state} not found in Historic Weather Database."
else:
df.set_index(0, inplace=True)
series = df.resample("MS").mean()
warnings.filterwarnings(
"ignore",
message=
"After 0.13 initialization must be handled at model creation")
result = ExponentialSmoothing(
series.astype(np.int64),
trend="add",
seasonal="add",
seasonal_periods=12).fit().forecast(24)
c = pd.Series([city.title()] * len(result.index))
c.index = result.index
s = pd.Series([state.upper()] * len(result.index))
s.index = result.index
result = pd.concat([c, s, result], axis=1)
result.columns = ["city", "state", "temp"]
result.index = result.index.astype(str)
# Converting for temperature in Fahrenheit if needed
# Conversion Helper Function
def to_fahr(temp: float) -> float:
return ((temp * 9) / 5) + 32
if metric != True:
result["temp"] = result["temp"].apply(to_fahr)
insert_pred = """
INSERT INTO {table}(
month,
city,
state,
mean
) VALUES%s
""".format(table=table)
execute_values(cur, insert_pred,
list(result.to_records(index=True)))
conn.commit()
result_json = result.to_json(indent=2)
conn.close()
return result_json
