Estimate future observation counts based on laji.fi data

forecast_species.py

import json
from statsmodels.tsa.holtwinters import Holt

# Load the JSON data
with open("pyjamalude-years.json", "r") as file:
    data = json.load(file)

'''
Estimate future observation counts based on laji.fi data
Made with GPT-4 code interpreter. Unvalidated.

Input: yearly aggregate of of observation count for a taxon.
e.g.
https://laji.fi/api/warehouse/query/unit/aggregate?target=MX.230561&countryId=ML.206&aggregateBy=gathering.conversions.year&selected=gathering.conversions.year&orderBy=gathering.conversions.year&page=1&pageSize=10000&onlyCount=false

'''
    
# Extract the results and organize the data
results = data["results"]
year_observation_data = [(int(item["aggregateBy"]["gathering.conversions.year"]), item["count"]) for item in results]
year_observation_data.sort(key=lambda x: x[0])
years, observations = zip(*year_observation_data)

# Fit the Holt's Linear Exponential Smoothing model
model = Holt(observations).fit()

# Forecast the next 7 years (2024-2030)
forecast = model.forecast(7)

forecast_years = list(range(2024, 2031))
forecast_values = list(forecast)

# Print the forecasted observations
for year, value in zip(forecast_years, forecast_values):
    print(f"{year}: Approximately {round(value)} observations")