def __init__(self, heirarchy, import_dir=None, export_dir=None, base_feature_count=None):
Exporter.__init__(self, export_dir=export_dir)
Importer.__init__(self, import_dir=import_dir)
self.tree = {}
self.dataset = []
self.heirarchy = heirarchy
if base_feature_count is None:
self.base_feature_count = 0
else:
self.base_feature_count = base_feature_count
self.max_depth = len(heirarchy)
def __init__(self, conditions, skip_sellingprice=True, import_dir=None, export_dir=None):
Importer.__init__(self, import_dir=import_dir)
Exporter.__init__(self, export_dir=export_dir)
self.conditions = conditions
self.skip_sellingprice = skip_sellingprice
# state: region
# e.g. ca: West
self.states = features.USA_MAP
# make: type
# e.g. honda: Normal
self.makes = features.MAKE_TYPE_MAP
# seller: type
# e.g. ford motor ca: Lease
self.sellers = features.SELLER_TYPE_MAP
self.dataset = []
self.filtered = []
self.output = []
try:
conditions[condition] = val['prices_centered']
except KeyError:
pass
results = []
for condition, offset in conditions.iteritems():
results.append([float(condition), np.average(offset)])
results = sorted(results, key = itemgetter(0))
# NOTE - there is clearly an issue with condition 4.0 selling prices :(
# it doesn't make any sense by that set of prices would be above the average
# while 3.9 and 4.1 are significantly below average
# let's just adjust this value and interpolate between 3.9 and 4.1
results[1][1] = (results[0][1] + results[2][1]) / 2
condition_map = {}
for item in results:
condition_map[str(item[0])] = item[1]
print item[0], item[1]
#plt.scatter([item[0] for item in results], [item[1] for item in results])
#plt.savefig('./plots/condition-offset.png')
EXPORT = Exporter()
EXPORT.save('condition-offset.json', dataset=condition_map)
try:
states[state] = states[state] + val['prices_centered']
except KeyError:
try:
states[state] = val['prices_centered']
except KeyError:
pass
results = []
for state, offset in states.iteritems():
count = len(offset)
if count < 1 / pow(confidence, 2):
print 'state: %s, n: %s' % (state, count)
print 'not enough data to calculate reliable offset'
else:
results.append([state, np.average(offset)])
results = sorted(results, key = itemgetter(0))
state_map = {}
for item in results:
state_map[item[0]] = item[1]
print item[0], item[1]
# plt.scatter([i for i, item in enumerate(results)], [item[1] for item in results])
# plt.savefig('./plots/state-offset.png')
EXPORT = Exporter()
EXPORT.save('state-offset.json', dataset=state_map)