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
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163 | class CellConnectionProbabilityEstimation(Component):
"""
Estimates the cell connection probabilities and posterior probabilities for each grid tile.
Cell connection probabilities are calculated based on footprint per grid.
Posterior probabilities are calculated based on the cell connection probabilities
and grid prior probabilities.
This class reads in cell footprint estimation and the grid model wit prior probabilities.
The output is a DataFrame that represents cell connection probabilities and
posterior probabilities for each cell and grid id combination for a given date.
"""
COMPONENT_ID = "CellConnectionProbabilityEstimation"
def __init__(self, general_config_path: str, component_config_path: str) -> None:
super().__init__(general_config_path, component_config_path)
self.data_period_start = datetime.datetime.strptime(
self.config.get(self.COMPONENT_ID, "data_period_start"), "%Y-%m-%d"
).date()
self.data_period_end = datetime.datetime.strptime(
self.config.get(self.COMPONENT_ID, "data_period_end"), "%Y-%m-%d"
).date()
self.data_period_dates = [
self.data_period_start + datetime.timedelta(days=i)
for i in range((self.data_period_end - self.data_period_start).days + 1)
]
self.current_date = None
self.current_cell_footprint = None
def initalize_data_objects(self):
self.clear_destination_directory = self.config.getboolean(self.COMPONENT_ID, "clear_destination_directory")
# Input
self.input_data_objects = {}
self.use_land_use_prior = self.config.getboolean(self.COMPONENT_ID, "use_land_use_prior")
inputs = {
"cell_footprint_data_silver": SilverCellFootprintDataObject,
}
if self.use_land_use_prior:
inputs["enriched_grid_data_silver"] = SilverEnrichedGridDataObject
for key, value in inputs.items():
path = self.config.get(CONFIG_SILVER_PATHS_KEY, key)
if check_if_data_path_exists(self.spark, path):
self.input_data_objects[value.ID] = value(self.spark, path)
else:
self.logger.warning(f"Expected path {path} to exist but it does not")
raise ValueError(f"Invalid path for {value.ID} in component {self.COMPONENT_ID} initialization")
# Output
self.output_data_objects = {}
silver_cell_probabilities_path = self.config.get(
CONFIG_SILVER_PATHS_KEY, "cell_connection_probabilities_data_silver"
)
if self.clear_destination_directory:
delete_file_or_folder(self.spark, silver_cell_probabilities_path)
self.output_data_objects[SilverCellConnectionProbabilitiesDataObject.ID] = (
SilverCellConnectionProbabilitiesDataObject(
self.spark,
silver_cell_probabilities_path,
)
)
@get_execution_stats
def execute(self):
self.logger.info(f"Starting {self.COMPONENT_ID}...")
self.read()
for current_date in self.data_period_dates:
self.logger.info(f"Processing cell footprint for {current_date.strftime('%Y-%m-%d')}")
self.current_date = current_date
self.current_cell_footprint = self.input_data_objects[SilverCellFootprintDataObject.ID].df.filter(
(F.make_date(F.col(ColNames.year), F.col(ColNames.month), F.col(ColNames.day)) == F.lit(current_date))
)
self.transform()
self.write()
self.spark.catalog.clearCache()
self.logger.info(f"Finished {self.COMPONENT_ID}")
def transform(self):
self.logger.info(f"Transform method {self.COMPONENT_ID}")
cell_footprint_df = self.current_cell_footprint
# Calculate the cell connection probabilities
window_spec = Window.partitionBy(ColNames.year, ColNames.month, ColNames.day, ColNames.grid_id)
cell_conn_probs_df = cell_footprint_df.withColumn(
ColNames.cell_connection_probability,
F.col(ColNames.signal_dominance) / F.sum(ColNames.signal_dominance).over(window_spec),
)
# Calculate the posterior probabilities
if self.use_land_use_prior:
grid_model_df = self.input_data_objects[SilverEnrichedGridDataObject.ID].df.select(
ColNames.grid_id, ColNames.prior_probability
)
cell_conn_probs_df = cell_conn_probs_df.join(grid_model_df, on=ColNames.grid_id)
cell_conn_probs_df = cell_conn_probs_df.withColumn(
ColNames.posterior_probability,
F.col(ColNames.cell_connection_probability) * F.col(ColNames.prior_probability),
)
elif not self.use_land_use_prior:
cell_conn_probs_df = cell_conn_probs_df.withColumn(
ColNames.posterior_probability,
F.col(ColNames.cell_connection_probability),
)
# Normalize the posterior probabilities
window_spec = Window.partitionBy(ColNames.year, ColNames.month, ColNames.day, ColNames.cell_id)
cell_conn_probs_df = cell_conn_probs_df.withColumn(
ColNames.posterior_probability,
F.col(ColNames.posterior_probability) / F.sum(ColNames.posterior_probability).over(window_spec),
)
cell_conn_probs_df = utils.apply_schema_casting(
cell_conn_probs_df, SilverCellConnectionProbabilitiesDataObject.SCHEMA
)
self.output_data_objects[SilverCellConnectionProbabilitiesDataObject.ID].df = cell_conn_probs_df
|