def run_all_hierarchi_clustering(rad,
date_range,
boxcox=True,
norm=True,
params=[
"bmnum", "noise.sky", "tfreq", "v", "p_l",
"w_l", "slist", "elv", "time_index"
],
methods=["agglomerative", "feature"],
n_clusters=[20, 5]):
"""
Invoke all partitioned clustering algorithm
rad: Radar code
date_range: Date range
"""
fd = FetchData(rad, date_range)
beams, _ = fd.fetch_data(
v_params=["elv", "v", "w_l", "gflg", "p_l", "slist", "v_e"])
rec = fd.convert_to_pandas(beams)
rec["time_index"] = utils.time_days_to_index(
[x.to_pydatetime() for x in rec["time"].tolist()])
if boxcox: rec = utils.boxcox_tx(rec)
if norm: rec = utils.normalize(rec, params)
print("\n", rec.head())
for _i, method in enumerate(methods):
print("\n >> Running {c} clustering".format(c=method))
model = Hierarchi(method,
rec[params].values,
n_clusters=n_clusters[_i])
model.setup()
model.run()
print("\n Estimating model skills.")
skill = Skills(model.data, model.obj.labels_)
return
def run_all_misc_clustering(
rad,
date_range,
boxcox=False,
norm=True,
params=["bmnum", "slist", "elv", "time_index", "v", "w_l"],
methods=["bang", "clique"]):
"""
Invoke all misc clustering algorithm
rad: Radar code
date_range: Date range
"""
fd = FetchData(rad, date_range)
beams, _ = fd.fetch_data(
v_params=["elv", "v", "w_l", "gflg", "p_l", "slist", "v_e"])
rec = fd.convert_to_pandas(beams)
rec["time_index"] = utils.time_days_to_index(
[x.to_pydatetime() for x in rec["time"].tolist()])
if boxcox: rec = utils.boxcox_tx(rec)
if norm: rec = utils.normalize(rec, params)
print("\n", rec.head())
for method in methods:
print("\n >> Running {c} clustering".format(c=method))
model = Misc(method, rec[params].values)
model.setup()
model.run()
print("\n Estimating model skills.")
skill = Skills(model.data, model.obj.labels_)
return
def run_all_densitybased_clustering(
rad,
date_range,
boxcox=False,
norm=True,
params=["bmnum", "v", "p_l", "w_l", "slist", "elv", "time_index"],
methods=["dbscan", "optics", "hdbscan"],
m_params={
"dbscan": {
"eps": 5.
},
"optics": {
"max_eps": 7.,
"metric": "minkowski"
},
"hdbscan": {
"metric": "minkowski",
"algorithm": "best"
}
}):
"""
Invoke all densitybased clustering algorithm
rad: Radar code
date_range: Date range
"""
fd = FetchData(rad, date_range)
beams, _ = fd.fetch_data(
v_params=["elv", "v", "w_l", "gflg", "p_l", "slist", "v_e"])
rec = fd.convert_to_pandas(beams)
rec["time_index"] = utils.time_days_to_index(
[x.to_pydatetime() for x in rec["time"].tolist()])
if boxcox: rec = utils.boxcox_tx(rec)
if norm: rec = utils.normalize(rec, params)
print("\n", rec.head())
for method in methods:
print("\n >> Running {c} clustering".format(c=method))
model = DBased(method, rec[params].values)
model.setup(m_params[method])
model.run()
print("\n Estimating model skills.")
skill = Skills(model.data, model.obj.labels_)
return
def plot_figure2(rad, bm, fname="figure2.png"):
files = glob.glob("../data/fitacf/%s/20021001*" % rad)
files.extend(glob.glob("../data/fitacf/%s/20021002*" % rad))
fd = FetchData(rad, [dt.datetime(2002, 10, 1),
dt.datetime(2002, 10, 3)],
files=files)
beams, _ = fd.fetch_data()
rec = fd.convert_to_pandas(beams)
rec = rec.sort_values(by=["time"])
print(" Total records: ", len(rec))
rti = RangeTimePlot(80, "", num_subplots=1)
rti.addPlot(rec,
bm,
param="v",
title="",
pmax=200,
step=50,
xlabel="Time UT")
rti.save("../figures/currie/" + fname)
return
def _ini_(self,
params=[
"bmnum", "noise.sky", "tfreq", "v", "p_l", "w_l", "slist",
"elv", "time_index"
],
v_params=["elv", "v", "w_l", "gflg", "p_l", "slist", "v_e"]):
"""
Model initialize
"""
print([self.stime, self.etime])
fd = FetchData(self.rad, [self.stime, self.etime])
beams, _ = fd.fetch_data(v_params=v_params)
self.rec = fd.convert_to_pandas(beams)
print(self.rec.head())
self.rec["time_index"] = utils.time_days_to_index(
[x.to_pydatetime() for x in self.rec["time"].tolist()])
self.rec["time"] = getD2N(self.rec["time"].tolist())
if hasattr(self, "boxcox") and self.boxcox:
self.rec = utils.boxcox_tx(self.rec)
if hasattr(self, "norm") and self.norm:
self.rec = utils.normalize(self.rec, params)
if self.verbose: print("\n", self.rec.head())
return