AtmoPredict/config.yaml at main · namanxdev/AtmoPredict · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
# Configuration for Extreme Weather Prediction System

# Data Collection
data:
  raw_data_path: "data/raw"
  processed_data_path: "data/processed"
  nasa_api_key: "FVVrb5hG30eYrxKxg91K4Jex1u8W75TKxBsMvYqL"  # Get from https://api.nasa.gov/

  # NASA POWER API endpoint for weather data
  power_api_url: "https://power.larc.nasa.gov/api/temporal/daily/point"

  # Date range for historical data
  start_date: "2010-01-01"
  end_date: "2023-12-31"

  # Variables to collect
  parameters:
    - "T2M"           # Temperature at 2 meters
    - "T2M_MAX"       # Maximum Temperature
    - "T2M_MIN"       # Minimum Temperature
    - "PRECTOTCORR"   # Precipitation
    - "WS2M"          # Wind Speed at 2 meters
    - "RH2M"          # Relative Humidity
    - "PS"            # Surface Pressure
    - "CLOUD_AMT"     # Cloud Amount

# Extreme Weather Thresholds
thresholds:
  very_hot:
    metric: "T2M_MAX"
    percentile: 95  # Top 5% of temperatures
    absolute: 35    # Or 35°C (95°F)

  very_cold:
    metric: "T2M_MIN"
    percentile: 5   # Bottom 5% of temperatures
    absolute: -5    # Or -5°C (23°F)

  very_windy:
    metric: "WS2M"
    percentile: 95
    absolute: 15    # 15 m/s (~33 mph)

  very_wet:
    metric: "PRECTOTCORR"
    percentile: 95
    absolute: 50    # 50mm rainfall

  very_uncomfortable:
    # Heat index or combination of temperature + humidity
    metric: "heat_index"
    percentile: 95
    absolute: 40

# Feature Engineering
features:
  rolling_window_days: [3, 7, 14, 30]
  lag_days: [1, 2, 3, 7]
  historical_comparison_years: 5

  calendar_features:
    - day_of_year
    - month
    - season
    - is_weekend

# Model Configuration
models:
  random_forest:
    n_estimators: 200
    max_depth: 15
    min_samples_split: 5
    class_weight: "balanced"

  xgboost:
    n_estimators: 300
    max_depth: 10
    learning_rate: 0.05
    scale_pos_weight: 3

  lightgbm:
    n_estimators: 300
    max_depth: 10
    learning_rate: 0.05
    num_leaves: 50

# Training Configuration
training:
  test_size: 0.2
  validation_size: 0.2
  random_seed: 42
  chronological_split: true
  cross_validation_folds: 5

# Evaluation Metrics
evaluation:
  metrics:
    - roc_auc
    - precision_recall_auc
    - brier_score
    - log_loss

  calibration:
    enabled: true
    n_bins: 10

# API Configuration
api:
  host: "127.0.0.1"
  port: 8081
  model_path: "models/trained"

# Weather API Configuration (for short-term forecasts)
weather_api:
  openweather_key: "your_openweather_api_key_here"  # Get from https://openweathermap.org/api
  # If not configured, system will use simulated data

# Frontend Configuration
frontend:
  title: "Extreme Weather Prediction System"
  default_location:
    latitude: 40.7128
    longitude: -74.0060
    name: "New York, NY"