Added single exponential smoothing API

int ndpi_ses_init(struct ndpi_ses_struct *ses, double alpha, float significance);
int ndpi_ses_add_value(struct ndpi_ses_struct *ses, const u_int32_t _value, double *forecast, double *confidence_band);
This commit is contained in:
Luca Deri
2021-03-11 00:04:33 +01:00
parent 5b7fe1360a
commit 6833ee2bbe
4 changed files with 235 additions and 29 deletions
+138 -14
View File
@@ -3722,20 +3722,53 @@ void rulesUnitTest() {
void rsiUnitTest() {
struct ndpi_rsi_struct s;
unsigned int v[] = {
31,
87,
173,
213,
223,
230,
238,
245,
251,
151,
259,
261,
264,
264,
270,
273,
288,
288,
304,
304,
350,
384,
423,
439,
445,
445,
445,
445
};
unsigned int v1[] = {
2227, 2219, 2208, 2217, 2218, 2213, 2223, 2243, 2224, 2229,
2215, 2239, 2238, 2261, 2336, 2405, 2375, 2383, 2395, 2363,
2382, 2387, 2365, 2319, 2310, 2333, 2268, 2310, 2240, 2217,
};
u_int i, n = sizeof(v) / sizeof(unsigned int);
u_int debug = 0;
assert(ndpi_alloc_rsi(&s, 8) == 0);
for(i=0; i<n; i++) {
float rsi = ndpi_rsi_add_value(&s, v[i]);
#if 0
printf("%2d) RSI = %f\n", i, rsi);
#endif
if(debug)
printf("%2d) RSI = %f\n", i, rsi);
}
ndpi_free_rsi(&s);
@@ -3796,7 +3829,6 @@ void hwUnitTest() {
void hwUnitTest2() {
struct ndpi_hw_struct hw;
u_int num_learning_points = 12;
u_int8_t trace = 1;
double v[] = {
31.908466339111,
@@ -3814,7 +3846,7 @@ void hwUnitTest2() {
264.78540039062,
264.78540039062,
270.47451782227,
273.3671875,
173.3671875,
288.34222412109,
288.34222412109,
304.24795532227,
@@ -3828,13 +3860,20 @@ void hwUnitTest2() {
445.05981445312,
445.05981445312
};
u_int num_learning_points = 1;
u_int i, num = sizeof(v) / sizeof(double);
float alpha = 0.5, beta = 0.5, gamma = 0.1;
assert(ndpi_hw_init(&hw, num_learning_points, 0 /* 0=multiplicative, 1=additive */, alpha, beta, gamma, 0.05) == 0);
float alpha = 0.9, beta = 0.5, gamma = 1;
FILE *fd = fopen("/tmp/result.csv", "w");
if(trace)
assert(ndpi_hw_init(&hw, num_learning_points, 0 /* 0=multiplicative, 1=additive */,
alpha, beta, gamma, 0.05) == 0);
if(trace) {
printf("\nHolt-Winters [alpha: %.1f][beta: %.1f][gamma: %.1f]\n", alpha, beta, gamma);
if(fd)
fprintf(fd, "index;value;prediction;lower;upper;anomaly\n");
}
for(i=0; i<num; i++) {
double prediction, confidence_band;
@@ -3843,13 +3882,94 @@ void hwUnitTest2() {
lower = prediction - confidence_band, upper = prediction + confidence_band;
if(trace)
if(trace) {
printf("%2u)\t%12.3f\t%.3f\t%12.3f\t%12.3f\t %s [%.3f]\n", i, v[i], prediction, lower, upper,
((rc == 0) || ((v[i] >= lower) && (v[i] <= upper))) ? "OK" : "ANOMALY",
confidence_band);
if(fd)
fprintf(fd, "%u;%.0f;%.0f;%.0f;%.0f;%s\n",
i, v[i], prediction, lower, upper,
((rc == 0) || ((v[i] >= lower) && (v[i] <= upper))) ? "OK" : "ANOMALY");
}
}
if(fd) fclose(fd);
ndpi_hw_free(&hw);
//exit(0);
}
/* *********************************************** */
void sesUnitTest() {
struct ndpi_ses_struct ses;
u_int8_t trace = 0;
double v[] = {
31.908466339111,
87.339714050293,
173.47660827637,
213.92568969727,
223.32124328613,
230.60134887695,
238.09457397461,
245.8137512207,
251.09228515625,
251.09228515625,
259.21997070312,
261.98754882812,
264.78540039062,
264.78540039062,
270.47451782227,
173.3671875,
288.34222412109,
288.34222412109,
304.24795532227,
304.24795532227,
350.92227172852,
384.54431152344,
423.25942993164,
439.43322753906,
445.05981445312,
445.05981445312,
445.05981445312,
445.05981445312
};
u_int num_learning_points = 1;
u_int i, num = sizeof(v) / sizeof(double);
float alpha = 0.9;
FILE *fd = fopen("/tmp/result.csv", "w");
assert(ndpi_ses_init(&ses, alpha, 0.05) == 0);
if(trace) {
printf("\nSingle Exponential Smoothing [alpha: %.1f]\n", alpha);
if(fd)
fprintf(fd, "index;value;prediction;lower;upper;anomaly\n");
}
ndpi_hw_free(&hw);
for(i=0; i<num; i++) {
double prediction, confidence_band;
double lower, upper;
int rc = ndpi_ses_add_value(&ses, v[i], &prediction, &confidence_band);
lower = prediction - confidence_band, upper = prediction + confidence_band;
if(trace) {
printf("%2u)\t%12.3f\t%.3f\t%12.3f\t%12.3f\t %s [%.3f]\n", i, v[i], prediction, lower, upper,
((rc == 0) || ((v[i] >= lower) && (v[i] <= upper))) ? "OK" : "ANOMALY",
confidence_band);
if(fd)
fprintf(fd, "%u;%.0f;%.0f;%.0f;%.0f;%s\n",
i, v[i], prediction, lower, upper,
((rc == 0) || ((v[i] >= lower) && (v[i] <= upper))) ? "OK" : "ANOMALY");
}
}
if(fd) fclose(fd);
}
/* *********************************************** */
@@ -3938,8 +4058,12 @@ int original_main(int argc, char **argv) {
return(-1);
}
// hwUnitTest2();
#ifdef HW_TEST
hwUnitTest2();
#endif
sesUnitTest();
/* Internal checks */
// binUnitTest();
//hwUnitTest();
+5
View File
@@ -1474,6 +1474,11 @@ extern "C" {
/* ******************************* */
int ndpi_ses_init(struct ndpi_ses_struct *ses, double alpha, float significance);
int ndpi_ses_add_value(struct ndpi_ses_struct *ses, const u_int32_t _value, double *forecast, double *confidence_band);
/* ******************************* */
int ndpi_jitter_init(struct ndpi_jitter_struct *hw, u_int16_t num_periods);
void ndpi_jitter_free(struct ndpi_jitter_struct *hw);
float ndpi_jitter_add_value(struct ndpi_jitter_struct *s, const float value);
+11
View File
@@ -1663,6 +1663,17 @@ struct ndpi_hw_struct {
double *s;
};
struct ndpi_ses_struct {
struct {
double alpha, ro;
} params;
u_int32_t num_values;
double sum_square_error, last_forecast, last_value;
};
/* **************************************** */
/* Prototype used to define custom DGA detection function */
typedef int (*ndpi_custom_dga_predict_fctn)(const char* domain, int domain_length);
+81 -15
View File
@@ -852,6 +852,8 @@ void ndpi_free_rsi(struct ndpi_rsi_struct *s) {
/* ************************************* */
// #define DEBUG_RSI
/*
This function adds a new value and returns the computed RSI, or -1
if there are too few points (< num_learning_values)
@@ -940,6 +942,7 @@ double ndpi_avg_inline(u_int32_t *v, u_int num) {
return(avg / (u_int32_t)num);
}
/* *********************************************************** */
/* *********************************************************** */
/*
@@ -951,7 +954,7 @@ double ndpi_avg_inline(u_int32_t *v, u_int num) {
additive If set to 1 will use the Holt-Winters additive seasonal (should be the default), otherwise the multiplicative seasonal.
alpha Level: specifies the coefficient for the level smoothing. Range 0..1. The higher α, the faster the method forgets old values
beta Trend: specifies the coefficient for the trend smoothing. Range 0..1.
gamma Seasonal: specifies the coefficient for the seasonal smoothing. Range 0..1. With gamma = 0, seaasonal correction is not used.
gamma Seasonal: specifies the coefficient for the seasonal smoothing. Range 0..1. With gamma = 0, seasonal correction is not used.
significance Significance level for the forecats sed for computing lower and upper bands. Range 0..1. Typical values 0.05 or less.
See https://en.wikipedia.org/wiki/Statistical_significance
@@ -978,8 +981,6 @@ int ndpi_hw_init(struct ndpi_hw_struct *hw,
if((significance < 0) || (significance > 1)) significance = 0.05;
hw->params.ro = ndpi_normal_cdf_inverse(1 - (significance / 2.));
hw->params.ro = 1.95996398454005;
if((hw->y = (u_int32_t*)ndpi_calloc(hw->params.num_season_periods, sizeof(u_int32_t))) == NULL)
return(-1);
@@ -1028,13 +1029,13 @@ int ndpi_hw_add_value(struct ndpi_hw_struct *hw, const u_int32_t _value, double
u_int idx = hw->num_values % hw->params.num_season_periods;
double prev_u, prev_v, prev_s, value = (double)_value;
double sq, error;
if(hw->num_values == hw->params.num_season_periods) {
double avg = ndpi_avg_inline(hw->y, hw->params.num_season_periods);
u_int i;
if(avg == 0) avg = 1; /* Avoid divisions by zero */
for(i=0; i<hw->params.num_season_periods; i++)
hw->s[i] = hw->y[i] / avg;
@@ -1043,7 +1044,7 @@ int ndpi_hw_add_value(struct ndpi_hw_struct *hw, const u_int32_t _value, double
hw->u = 0;
else
hw->u = _value / hw->s[i];
hw->v = 0;
ndpi_free(hw->y);
hw->y = NULL;
@@ -1051,12 +1052,12 @@ int ndpi_hw_add_value(struct ndpi_hw_struct *hw, const u_int32_t _value, double
idx = hw->num_values % hw->params.num_season_periods;
prev_u = hw->u, prev_v = hw->v, prev_s = hw->s[idx];
if(prev_s != 0)
hw->u = ((hw->params.alpha * value) / prev_s) + ( 1 - hw->params.alpha) * (hw->u + hw->v);
else
hw->u = 0; /* Avoid divisions by zero */
hw->v = (hw->params.beta * (hw->u - prev_u)) + ((1 - hw->params.beta ) * hw->v);
if(hw->u != 0)
@@ -1080,7 +1081,7 @@ int ndpi_hw_add_value(struct ndpi_hw_struct *hw, const u_int32_t _value, double
*forecast, hw->sum_square_error,
sq, *confidence_band);
#endif
hw->num_values++, idx = (idx + 1) % hw->params.num_season_periods;
return(1); /* We're in business: forecast is meaningful now */
@@ -1092,7 +1093,7 @@ int ndpi_hw_add_value(struct ndpi_hw_struct *hw, const u_int32_t _value, double
/*
Jitter calculator
Used to determine how noisy is a signal
*/
@@ -1100,7 +1101,7 @@ int ndpi_jitter_init(struct ndpi_jitter_struct *s, u_int16_t num_learning_values
memset(s, 0, sizeof(struct ndpi_jitter_struct));
if(num_learning_values < 2) num_learning_values = 2;
s->empty = 1, s->num_values = num_learning_values;
s->observations = (float*)ndpi_calloc(num_learning_values, sizeof(float));
@@ -1108,7 +1109,7 @@ int ndpi_jitter_init(struct ndpi_jitter_struct *s, u_int16_t num_learning_values
s->last_value = 0;
return(0);
} else
return(-1);
return(-1);
}
/* ************************************* */
@@ -1132,7 +1133,7 @@ float ndpi_jitter_add_value(struct ndpi_jitter_struct *s, const float value) {
s->observations[s->next_index] = val;
s->jitter_total += val;
}
s->last_value = value, s->next_index = (s->next_index + 1) % s->num_values;
if(s->next_index == 0) s->jitter_ready = 1; /* We have completed one round */
@@ -1141,10 +1142,75 @@ float ndpi_jitter_add_value(struct ndpi_jitter_struct *s, const float value) {
value, val, s->jitter_total,
s->jitter_ready ? (s->jitter_total / s->num_values) : -1);
#endif
if(!s->jitter_ready)
return(-1); /* Too early */
else
else
return(s->jitter_total / s->num_values);
}
/* *********************************************************** */
/* *********************************************************** */
/*
Single Exponential Smoothing
*/
int ndpi_ses_init(struct ndpi_ses_struct *ses, double alpha, float significance) {
memset(ses, 0, sizeof(struct ndpi_ses_struct));
ses->params.alpha = alpha;
if((significance < 0) || (significance > 1)) significance = 0.05;
ses->params.ro = ndpi_normal_cdf_inverse(1 - (significance / 2.));
return(0);
}
/* *********************************************************** */
/*
Returns the forecast and the band (forecast +/- band are the upper and lower values)
Input
ses: Datastructure previously initialized
value The value to add to the measurement
Output
forecast The forecasted value
confidence_band The value +/- on which the value should fall is not an anomaly
Return code
0 Too early: we're still in the learning phase. Output values are zero.
1 Normal processing: forecast and confidence_band are meaningful
*/
int ndpi_ses_add_value(struct ndpi_ses_struct *ses, const u_int32_t _value, double *forecast, double *confidence_band) {
double value = (double)_value, error;
int rc;
if(ses->num_values == 0)
*forecast = value;
else
*forecast = (ses->params.alpha * (ses->last_value - ses->last_forecast)) + ses->last_forecast;
error = value - *forecast;
ses->sum_square_error += error * error;
if(ses->num_values > 0) {
double sq = sqrt(ses->sum_square_error / (ses->num_values+1));
*confidence_band = ses->params.ro * sq;
rc = 1;
} else
*confidence_band = 0, rc = 0;
ses->num_values++, ses->last_value = value, ses->last_forecast = *forecast;
#ifdef SES_DEBUG
printf("[num_values: %u][[error: %.3f][forecast: %.3f][sqe: %.3f][sq: %.3f][confidence_band: %.3f]\n",
ses->num_values, error, *forecast, ses->sum_square_error, sq, *confidence_band);
#endif
return(rc);
}