feat(tf): transform dataframe to tensorflow dataset

constants.py

@@ -0,0 +1,18 @@
+import os
+
+directory = os.path.dirname(os.path.abspath(__file__))
+
+GEODATA = os.path.join(directory, 'geodata')
+ECOREGIONS = os.path.join(GEODATA, 'ecoregions', 'single-parts.shp')
+ELEVATION = os.path.join(GEODATA, 'srtm', 'topo30-180.tif')
+TEMP = os.path.join(GEODATA, 'air_temp')
+PRECIP = os.path.join(GEODATA, 'precipitation')
+
+MIN_YEAR = 1900
+MAX_YEAR = 2017
+
+SEASONS = ['winter', 'spring', 'summer', 'autumn']
+WINTER_MONTHS = ['december', 'january', 'february']
+SPRING_MONTHS = ['march', 'april', 'may']
+SUMMER_MONTHS = ['june', 'july', 'august']
+AUTUMN_MONTHS = ['september', 'november', 'october']

data.py

@@ -1,19 +1,12 @@
 import geopandas
-import os
 import rasterio
 import pandas as pd
 import numpy as np
 import time
 from matplotlib import pyplot
 from shapely.geometry import Point
+from constants import *
 
-directory = os.path.dirname(os.path.abspath(__file__))
-
-GEODATA = os.path.join(directory, 'geodata')
-ECOREGIONS = os.path.join(GEODATA, 'ecoregions', 'single-parts.shp')
-ELEVATION = os.path.join(GEODATA, 'srtm', 'topo30-180.tif')
-TEMP = os.path.join(GEODATA, 'air_temp')
-PRECIP = os.path.join(GEODATA, 'precipitation')
 
 def read_temp_data(year):
     return pd.read_csv(os.path.join(TEMP, 'air_temp.{}'.format(year)), sep='\s+', header=None,
@@ -38,8 +31,6 @@ elevation_data = elevation.read(1)
 
 temp = {}
 precip = {}
-MIN_YEAR = 1900
-MAX_YEAR = 2017
 for year in range(MIN_YEAR, MAX_YEAR + 1):
     temp[year] = read_temp_data(year)
     precip[year] = read_precip_data(year)
@@ -52,7 +43,7 @@ boundary = world.boundary
 temp_precip_columns = []
 
 for year in range(MIN_YEAR, MAX_YEAR + 1):
-    for s in ['winter', 'spring', 'summer', 'autumn']:
+    for s in SEASONS:
         temp_precip_columns += ['temp_{}_{}'.format(s, year), 'precip_{}_{}'.format(s, year)]
 
 columns = ['biome_num', 'biome_name', 'elevation', 'distance_to_water'] + temp_precip_columns
@@ -86,14 +77,14 @@ def get_point_information(longitude, latitude):
         winter_temp = [yt.january, yt.february] + ([yearly_temp[year - 1].december] if year > MIN_YEAR else [])
         winter_precip = [yp.january, yp.february] + ([yearly_precip[year - 1].december] if year > MIN_YEAR else [])
 
-        spring_temp = [yt[month] for month in ['march', 'april', 'may']]
-        spring_precip = [yp[month] for month in ['march', 'april', 'may']]
+        spring_temp = [yt[month] for month in SPRING_MONTHS]
+        spring_precip = [yp[month] for month in SPRING_MONTHS]
 
-        summer_temp = [yt[month] for month in ['june', 'july', 'august']]
-        summer_precip = [yp[month] for month in ['june', 'july', 'august']]
+        summer_temp = [yt[month] for month in SUMMER_MONTHS]
+        summer_precip = [yp[month] for month in SUMMER_MONTHS]
 
-        autumn_temp = [yt[month] for month in ['september', 'november', 'october']]
-        autumn_precip = [yp[month] for month in ['september', 'november', 'october']]
+        autumn_temp = [yt[month] for month in AUTUMN_MONTHS]
+        autumn_precip = [yp[month] for month in AUTUMN_MONTHS]
 
         item['temp_winter_{}'.format(year)] = np.mean(winter_temp)
         item['precip_winter_{}'.format(year)] = np.mean(winter_temp)

nn.py

@@ -7,7 +7,18 @@ from tensorflow import keras
 # Helper libraries
 import numpy as np
 import matplotlib.pyplot as plt
+import pandas as pd
 
-import data
+from utils import *
+
+tf.enable_eager_execution()
+
+df = pd.read_pickle('data_distance.p')
+# print(df.head())
+
+dataset = dataframe_to_dataset_biomes(df)
+
+for feature, target in dataset:
+    print('{} => {}'.format(feature, target))
 
 print(tf.__version__)

utils.py

@@ -0,0 +1,39 @@
+import numpy as np
+import tensorflow as tf
+import pandas as pd
+from constants import *
+
+inputs = ['elevation', 'distance_to_water']
+output = 'biome_num'
+
+def dataframe_to_dataset_biomes(df):
+    rows = df.shape[0]
+
+    # 8 for seasonal temp and precipitation
+    # 3 for latitude, elevation and distance_to_water
+    columns = 11 
+
+    tf_inputs = np.empty((0, columns))
+    tf_output = np.empty((0))
+    latitude = np.array(df.index.get_level_values(1))
+    longitude = np.array(df.index.get_level_values(0))
+
+    for year in range(MIN_YEAR, MAX_YEAR + 1):
+        local_inputs = list(inputs)
+        for season in SEASONS:
+            local_inputs += [
+                'temp_{}_{}'.format(season, year),
+                'precip_{}_{}'.format(season, year)
+            ]
+
+
+        local_df = df[local_inputs]
+        local_df.loc[:, 'latitude'] = pd.Series(latitude, index=local_df.index)
+
+        tf_inputs = np.concatenate((tf_inputs, local_df.values), axis=0)
+        tf_output = np.concatenate((tf_output, df[output].values), axis=0)
+
+    tf_inputs = tf.cast(tf_inputs, tf.float32)
+    tf_output = tf.cast(tf_output, tf.int32)
+
+    return tf.data.Dataset.from_tensor_slices((tf_inputs, tf_output))