pandorafms/pandora_console/include/functions_forecast.php

<?php

// Pandora FMS - http://pandorafms.com
// ==================================================
// Copyright (c) 2005-2010 Artica Soluciones Tecnologicas
// Please see http://pandorafms.org for full contribution list

// This program is free software; you can redistribute it and/or
// modify it under the terms of the  GNU Lesser General Public License
// as published by the Free Software Foundation; version 2

// This program is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
// GNU General Public License for more details.

/**
 * @package Include
 * @subpackage Forecast
 */

/**
 * Create a prediction based on module data with least square method (linear regression) 
 *
 * @param int Module id.
 * @param int Period of the module data.
 * @param int Period of the prediction or false to use it in prediction_date function (see below). 
 * @param int Maximun value using this function for prediction_date.
 * @param int Minimun value using this function for prediction_date.
 * @param bool Result data for CSV file exportation.
 * 
 * @return array Void array or prediction of the module data.
 */
function forecast_projection_graph($module_id,
	$period = SECONDS_2MONTHS, $prediction_period, $max_value = false,
	$min_value = false, $csv = false) {
	
	global $config;
	
	$max_exec_time = ini_get('max_execution_time'); 
	
	if ($max_exec_time !== false) {
		
		$max_exec_time = (int)$max_exec_time; 
	
	}
	
	$begin_time = time();

	$params =array(
		'agent_module_id'     => $module_id,
		'period'              => $period,
		'return_data'         => 1,
		'projection'          => true
	);

	$module_data = grafico_modulo_sparse ($params);
	
	if (empty($module_data)) {
		return array();	
	}
	// Prevents bad behaviour over image error 
	else if (!is_array($module_data) and preg_match('/^<img(.)*$/', $module_data)) {
		return;
	}
	
	// Data initialization
	$sum_obs        = 0;
	$sum_xi         = 0;
	$sum_yi         = 0;
	$sum_xi_yi      = 0;
	$sum_xi2        = 0;
	$sum_yi2        = 0;
	$sum_diff_dates = 0;
	$last_timestamp = get_system_time();
	$agent_interval = SECONDS_5MINUTES;
	$cont           = 1;
	$data           = array();
	//$table->data = array();

	// Creates data for calculation
	if (is_array($module_data) || is_object($module_data)) {
		foreach ($module_data['sum1']['data'] as $key => $row) {
			if ($row[0] == '') {
				continue;
			}

			$row[0] = $row[0] / 1000;

			$data[0] = '';
			$data[1] = $cont;
			$data[2] = date($config["date_format"], $row[0]);
			$data[3] = $row[0];
			$data[4] = $row[1];
			$data[5] = $row[0] * $row[1];
			$data[6] = $row[0] * $row[0];
			$data[7] = $row[1] * $row[1];
			if ($cont == 1) {
				$data[8] = 0;
			}
			else {
				$data[8] = $row[0] - $last_timestamp;
			}

			$sum_obs        = $sum_obs + $cont;
			$sum_xi         = $sum_xi + $row[0];
			$sum_yi         = $sum_yi + $row[1];
			$sum_xi_yi      = $sum_xi_yi + $data[5];
			$sum_xi2        = $sum_xi2 + $data[6];
			$sum_yi2        = $sum_yi2 + $data[7];
			$sum_diff_dates = $sum_diff_dates + $data[8];
			$last_timestamp = $row[0];
			$cont++;
		}
	}

	$cont--;

	// Calculation over data above:
	// 1. Calculation of linear correlation coefficient...
	
	// 1.1 Average for X: Sum(Xi)/Obs 
	// 1.2 Average for Y: Sum(Yi)/Obs
	// 2. Covariance between vars
	// 3.1  Standard deviation for X: sqrt((Sum(Xi²)/Obs) - (avg X)²) 
	// 3.2 Standard deviation for Y: sqrt((Sum(Yi²)/Obs) - (avg Y)²) 
	// Linear correlation coefficient:
	
		
/*
	if ($cont != 0) {
		$covariance = $sum_xi_yi/$cont;
		$dev_x = sqrt(($sum_xi2/$cont) - ($avg_x*$avg_x));
		$dev_y = sqrt(($sum_yi2/$cont) - ($avg_y*$avg_y));
	} else {
		$covariance = 0;
		$dev_x = 0;
		$dev_y = 0;
	}
	// Prevents division by zero
	if ($dev_x != 0 and $dev_y != 0) {
		$linear_coef = $covariance / ($dev_x * $dev_y);
	}
*/
	// Agent interval could be zero, 300 is the predefined
	if ($sum_obs == 0) {
		$agent_interval = SECONDS_5MINUTES;
	}
	else {
		$agent_interval =  $sum_diff_dates / $sum_obs;
	}
	
	// Could be a inverse correlation coefficient
	// if $linear_coef < 0.0
	//	 if $linear_coef >= -1.0 and $linear_coef <= -0.8999
	// 		Function variables have an inverse linear relathionship!
	//   else 
	// 		Function variables don't have an inverse linear relathionship!	
	
	// Could be a direct correlation coefficient
	// else 
	//	 if ($linear_coef >= 0.8999 and $linear_coef <= 1.0) {
	//		Function variables have a direct linear relathionship!
	//   else 
	//		Function variables don't have a direct linear relathionship!
	
	// 2. Calculation of linear regresion...
	
	$b_num = (($cont * $sum_xi_yi) - ($sum_xi * $sum_yi));
	$b_den = (($cont * $sum_xi2) - ($sum_xi * $sum_xi));
	if ($b_den == 0)
		return;
	$b = $b_num / $b_den;
	
	$a_num = ($sum_yi) - ($b * $sum_xi);
	
	if ($cont != 0) {
		$a = $a_num / $cont;
	} else {
		$a = 0;
	}
	
	// Data inicialization
	$output_data = array();
	if ($prediction_period != false) {
		$limit_timestamp = $last_timestamp + $prediction_period;
	}
	$current_ts = $last_timestamp;
	$in_range = true;
	$time_format_2 = '';
	
	$temp_range = $period;
	if ($period < $prediction_period)
		$temp_range = $prediction_period;
	
	if ($temp_range <= SECONDS_6HOURS) {
		$time_format = 'H:i:s';
	}
	elseif ($temp_range < SECONDS_1DAY) {
		$time_format = 'H:i';
	}
	elseif ($temp_range < SECONDS_15DAYS) {
		$time_format = 'M d';
		$time_format_2 = 'H\h';
	}
	elseif ($temp_range <= SECONDS_1MONTH) {
		$time_format = 'M d';
		$time_format_2 = 'H\h';
	} 
	else {
		$time_format = 'M d';
	}
	
	// Aplying linear regression to module data in order to do the prediction	
	$output_data = array();
	$idx = 0;
	// Create data in graph format like
	while ($in_range) {
		$now = time();

		//  Check that exec time is not greater than half max exec server time
		if ($max_exec_time !== false) {
			if (($begin_time + ($max_exec_time/2)) < $now) {
				return false;
			}
		}

		$timestamp_f = $current_ts * 1000;

		if ($csv) {
			$output_data[$idx]['date'] = $current_ts;
			$output_data[$idx]['data'] = ($a + ($b * $current_ts));
		}
		else {
			$output_data[$idx][0] = $timestamp_f;
			$output_data[$idx][1] = ($a + ($b * $current_ts));
		}

		// Using this function for prediction_date
		if ($prediction_period == false) {
			// These statements stop the prediction when interval is greater than 2 years
			if ($current_ts - $last_timestamp >= 94608000) {
				return false;
			}
			
			// Found it
			if ($max_value >= $output_data[$idx][0] and $min_value <= $output_data[$idx][0]) {
				return $current_ts;
			}
		}
		else if ($current_ts > $limit_timestamp) {
			$in_range = false;
		}
		$current_ts = $current_ts + $agent_interval;
		$idx++;
	}

	return $output_data;
}

/**
 * Return a date when the date interval is reached
 *
 * @param int Module id.
 * @param int Given data period to make the prediction
 * @param int Max value in the interval.
 * @param int Min value in the interval. 
 * 
 * @return mixed timestamp with the prediction date or false
 */
function forecast_prediction_date ($module_id,
	$period = SECONDS_2MONTHS, $max_value = 0, $min_value = 0) {
	// Checks interval
	if ($min_value > $max_value) {
		return false;
	}
	
	return forecast_projection_graph($module_id, $period, false, $max_value, $min_value);
}