Method for recognizing multi-dimensional anomalous urban traffic event based on ternary gaussian mixture model
A method for recognizing multi-dimensional anomalous urban traffic events based on a ternary Gaussian mixture model includes: reading a data sample of urban road traffic events; randomly dividing the data sample into a first subsample and a second subsample; performing modeling based on the first subsample by using the ternary Gaussian mixture model to obtain a second ternary Gaussian mixture model to calculate a distribution probability p of any sample point; clustering the second subsample, recognizing an outlier in the second subsample, and labeling the outlier and a normal point to obtain a labeled subsample; calculating the labeled subsample to obtain the distribution probability p corresponding to each sample point in the labeled subsample; when a new traffic event occurs, obtaining features of three dimensions of the new traffic event, calculating a distribution probability p by using the second model, and recognizing the new traffic event as anomalous if p<t-score.
1. A method for recognizing multi-dimensional anomalous urban traffic events based on a second ternary Gaussian mixture model, comprising the following steps:
step 1: reading a data sample of urban road traffic events using processor, wherein an input of the data sample comprises features of three dimensions comprising a traffic event quantity based on an event sequence, a weather condition, and a traffic congestion index;
step 2: randomly dividing the data sample into a first subsample and a second subsample;
step 3: performing a modeling using based on the first subsample using the processor by using a first ternary Gaussian mixture model to obtain the second ternary Gaussian mixture model, wherein the second ternary Gaussian mixture model is configured to calculate a distribution probability p of any sample point;
step 4: clustering the second subsample by using a density-based spatial clustering of applications with noise (DBSCAN) algorithm, recognizing a first outlier in the second subsample, and labeling the first outlier and a normal point to change the second subsample to a labeled subsample;
step 5: calculating the labeled subsample by using the second ternary gaussian mixture model obtained using the processor in step 3 to obtain the distribution probability p corresponding to each sample point x in the labeled subsample, wherein a distribution probability p allowing F1score to reach a maximum is a threshold t-score, and F1score is calculated by the following formula:
F
1
score
=
2
·
precision
·
recall
precision
+
recall
,
wherein
precision
=
tp
tp
+
fp
,
recall
=
tp
tp
+
fn
,
tp represents a quantity of true-positive sample points, fp represents a quantity of false-positive sample points, fn represents a quantity of false-negative sample points, the true-positive sample points are defined as first sample points with both a first anomalous model prediction result and a first anomalous actual result, the false-positive sample points are defined as second sample points with a second anomalous model prediction result and a normal actual result, and the false-negative sample points are defined as third sample points with a normal model prediction result and a second anomalous actual result, wherein a method of selecting the threshold t-score comprises the following steps:
step 501 : initializing an initial value of p′ and a highest value best_f1 of F1score as 0, and selecting a step, wherein step=(max(P 3 )−min(P 3 ))/1000, wherein P 3 represents a set of the distribution probability p corresponding to the each sample point x in the labeled subsample;
step 502 : setting a value of p 1 ′ to a sum of a minimum value in the P 3 and one step to obtain p 1 ′=min(P 3 )+step;
step 503 : extracting a sample point with a distribution probability p less than p 1 ′ from the labeled subsample, determining, by using the second ternary Gaussian mixture model, that the sample point is a second outlier, calculating the F1score, and denoting a calculated value as f1;
step 504 : comparing f1 and best_f1; if f1 is greater than best_f1, setting a value of best_f1 to f1, and assigning a value of p 1 ′ to p′ to obtain p′=p 1 ; and if f1 is not greater than best_f1, keeping the value of best_f1 and a value of p′ unchanged; and
step 505 : repeating steps 502 to 504 cyclically, and increasing p 1 ′ by one step each time until p 1 ′=max (P 3 ), wherein
a final value of p′ is the threshold t-score of an anomalous event on an urban road section; and
step 6: when a new traffic event occurs, obtaining features of three dimensions of the new traffic event, calculating a distribution probability p by using the second ternary Gaussian mixture model, and recognizing the new traffic event as anomalous if p<t-score.
2. The method for recognizing the multi-dimensional anomalous urban traffic events based on the second ternary Gaussian mixture model according to claim 1 , wherein
in step 2, a ratio of the first subsample to the second subsample is 9:1.
3. The method for recognizing the multi-dimensional anomalous urban traffic events based on the second ternary Gaussian mixture model according to claim 1 , wherein
in step 5, the distribution probability p is calculated by the following formula:
p
(
x
;
μ
,
∑
)
=
1
(
2
π
)
n
/
2
∑
1
/
2
exp
(
-
1
2
(
x
-
μ
)
T
∑
-
1
(
x
-
μ
)
)
;
wherein p (x; μ, Σ) represents a distribution probability of a sample point x in the labeled subsample, μ represents a mean vector of each dimension in the labeled subsample, μ=[μ 1 , μ 2 , μ 3 ], μ 1 represents a mean value of the traffic event quantity, μ 2 represents a mean value of the weather condition, μ 3 represents a mean value of the traffic congestion index, Σ represents a covariance matrix of the each dimension in the labeled subsample,
∑
=
σ
1
2
0
0
0
σ
2
2
0
0
0
σ
3
2
,
σ 1 represents a standard deviation of the traffic event quantity, σ 2 represents a standard deviation of the weather condition, and σ 3 represents a standard deviation of the traffic congestion index.