Due to the complexity and little knowledge of design objects, both the quality criteria and the limitations of the parametric optimization problem are, as a rule, too complex for the use of classical methods of searching for an extremum. Therefore, in practice, preference is given to search engine optimization methods. Let's consider the main stages of any search method.
The initial data in the search methods are the required accuracy of the method e and the starting point of the search X 0 .
Then the search step size is selected h, and according to some rule new points are obtained X k +1 by previous point X k at k= 0, 1, 2, … Obtaining new points continues until the condition for stopping the search is met. The last search point is considered the solution to the optimization problem. All search points make up the search trajectory.
Search methods differ from each other in the procedure for choosing the step size h(the step can be the same at all iterations of the method or calculated at each iteration), the algorithm for obtaining a new point and the condition for stopping the search.
For methods using a constant step size, h much less precision should be chosen e. If at the selected step size h If it is not possible to obtain a solution with the required accuracy, then you need to reduce the step size and continue the search from the last point of the existing trajectory.
It is customary to use the following as search termination conditions:
1) all neighboring search points are worse than the previous one;
2) ç F(X k +1 )–F(X k ) ç £ e, that is, the values of the objective function F(X) at adjacent points (new and previous) differ from each other by an amount no more than the required accuracy e;
3) ,i = 1, …, n, that is, all partial derivatives at the new search point are practically equal to 0, that is, they differ from 0 by an amount not exceeding the accuracy e.
Algorithm for obtaining a new search point X k+1 on previous point X k different for each search method, but each new search point must be no worse than the previous one: if the optimization problem is a minimum search problem, then F(X k +1 ) £ F(X k ).
Search engine optimization methods are usually classified according to the order of the derivative of the objective function used to obtain new points. Thus, in zero-order search methods, the calculation of derivatives is not required, but the function itself is sufficient F(X). First order search methods use the first partial derivatives, and second order methods use the second derivative matrix (Hessian matrix).
The higher the order of the derivatives, the more justified the choice of a new search point and the lower the number of iterations of the method. But at the same time, each iteration is labor intensive due to the need for numerical calculation of derivatives.
The effectiveness of the search method is determined by the number of iterations and the number of calculations of the objective function F(X) at each iteration of the method.
Let's consider most common search methods, arranging them in order of decreasing number of iterations.
For zero order search methods the following is true: in the random search method it is impossible to predict in advance the number of calculations F(X) in one iteration N, and in the coordinate descent method N£2× n, Where n- number of controlled parameters X = (x 1 , x 2 .,…, x n ).
For first order search methods the following estimates are valid: in the gradient method with a constant step N = 2 × n; in the gradient method with step division N=2 × n + n 1 , Where n 1 – number of calculations F(X), necessary for checking the step crushing conditions; in the steepest descent method N = 2 × n + n 2 , Where n 2 – number of calculations F(X), necessary for calculating the optimal step size; and in the Davidon-Fletcher-Powell (DFP) method N = 2 × n + n 3 , Where n 3 – number of calculations F(X), necessary to calculate a matrix approximating the Hessian matrix (for quantities n 1 , n 2 , n 3 the ratio is valid n 1 < n 2 < n 3 ).
And finally in the second order method- Newton's method N = 3 × n 2 .
When obtaining these estimates, an approximate calculation of derivatives using finite difference formulas is assumed, that is, to calculate the first-order derivative, two values of the objective function are needed F(X), and for the second derivative – the values of the function at three points.
In practice, the steepest descent method and the DFT method have found widespread use as methods with an optimal ratio of the number of iterations and their labor intensity.
Let's start looking at zero order search methods. In the random search method, the initial data is the required accuracy of the method e, the starting point of the search X 0 = (x 1 0 , x 2 0 , …, x n 0 ) and search step size h.
The search for new points is carried out in a random direction, on which the specified step is postponed h, thus obtaining a trial point 
and checking whether the trial point is better than the previous search point. For the minimum search problem this means that:
(6.19)
If this condition is met, then the trial point is included in the search trajectory ( 
). Otherwise, the trial point is excluded from consideration and a new random direction from the point is selected X k ,
k= 0, 1, 2, … (Fig. 6.3).
X k +1
F(X)
Despite the simplicity of this method, its main disadvantage is the fact that it is not known in advance how many random directions will be required to obtain a new point on the search trajectory X k +1 , which makes the cost of one iteration too high.
Rice. 6.3. Towards the random search method
In addition, since choosing the search direction does not use information about the objective function F(X), the number of iterations in the random search method is very large.
In this regard, the random search method is used to study little-studied design objects and to escape the zone of attraction of the local minimum when searching for the global extremum of the objective function.
Unlike the random search method, in the coordinate descent method, directions parallel to the coordinate axes are chosen as possible search directions, and movement is possible both in the direction of increasing and decreasing the coordinate value.
The initial data in the coordinate descent method is the step size h and the starting point of the search X 0
= (x 1
0
,
x 2
.
0
,…,
x n 0
)
. We start the movement from the point X 0
along the x 1 axis in the direction of increasing coordinates. Let's get a test point 
(x 1
k +
h,
x 2
k ,…,
x n k),
k= 0. Let's compare the value of the function F(X) with the value of the function at the previous search point X k .
If 
(we assume that we need to solve the minimization problem F(X), then the trial point is included in the search trajectory ( 
)
.
Otherwise, we exclude the trial point from consideration and obtain a new trial point, moving along the axis x 1
towards decreasing coordinates. Let's get a test point 
(x 1
k –
h,
x 2
k ,…,
x n k). Check if 
, then we continue moving along the x 2 axis in the direction of increasing coordinates. Let's get a test point 
(x 1
k +
h,
x 2
k ,…,
x n k), etc.
When constructing a search trajectory, repeated movement along points included in the search trajectory is prohibited.
Obtaining new points in the coordinate descent method continues until point X k is obtained for which all neighboring 2× n trial points (in all directions x 1
,
x 2
,
…,
x n in the direction of increasing and decreasing coordinate values) will be worse, that is 
. Then the search stops and the last point of the search trajectory is selected as the minimum point X*= X k .
Let's look at the operation of the coordinate descent method using an example (Fig. 2.21): n = 2, X = (x 1 , x 2 ), F (x 1 , x 2 ) min, F(x 1 , x 2 ) = (x 1 – 1) 2 + (x 2 – 2) 2 , h= 1, X 0 = (0, 1) .
We begin to move along the axis x 1 upward
coordinates. Let's get the first test point
(x 1
0
+
h,
x 2
0
) = (1, 1), F(
)
= (1-1) 2 +
(1-2) 2 =
1,
F(X 0 ) = (0-1) 2 + (1-2) 2 = 2,
F( 
)
< Ф(Х
0
)
X 1
= (1, 1).
x 1 from point X 1
=(x 1
1
+
h,
x 2
1
) = (2, 1), F( 
)
= (2-1) 2 +
(1-2) 2 =
2,
F(X 1 ) = (1-1) 2 + (1-2) 2 = 1,
that is F( 
) > Ф(Х 1
)
– the trial point with coordinates (2, 1) is excluded from consideration, and the search for the minimum continues from the point X 1
.
We continue moving along the axis x 2 from point X 1 towards increasing coordinates. Let's get a test point
= (x 1
1
,
x 2
1
+
h)
= (1, 2), F( 
)
= (1-1) 2
+ (2-2) 2
= 0,
F(X 1 ) = (1-1) 2 + (1-2) 2 = 1,
F( 
)
< Ф(Х
1
)
X 2
= (1,
2).
We continue moving along the axis x 2 from point X 2 towards increasing coordinates. Let's get a test point
= (x 1
2
,
x 2
2
+
h)
= (1, 3), F( 
)
= (1-1) 2 +
(3-2) 2 =
1,
F(X 2 ) = (1-1) 2 + (2-2) 2 = 0,
that is F( 
) > Ф(Х 2
)
– the trial point with coordinates (1, 3) is excluded from consideration, and the search for the minimum continues from the point X 2
.
5. Continue moving along the axis x 1 from point X 2 towards increasing coordinates. Let's get a test point
= (x 1
2
+
h,
x 2
2
) = (2, 2), F( 
)
= (2-1) 2 +
(2-2) 2 =1,
F(X 2 ) = (1-1) 2 + (2 - 2) 2 = 0,
that is F(X ^ ) > Ф(Х 2 ) – the trial point with coordinates (2, 2) is excluded from consideration, and the search for the minimum continues from the point X 2 .
6. Continue moving along the axis x 1 from point X 2 towards decreasing coordinates. Let's get a test point
= (x 1
2
-
h,
x 2
2
)
= (0, 2), F( 
)
= (0-1) 2 +(2-2) 2
= 1,
F(X 2 ) = (1-1) 2 + (2 - 2) 2 = 0,
that is F( 
) > Ф(Х 2
)
– the trial point with coordinates (0, 2) is excluded from consideration, and the search for the minimum is completed, since for the point X 2
The condition for stopping the search is met. Minimum point of the function F(x 1
,
x 2
)
= (x 1
– 1) 2 +
(x 2
– 2) 2 is X *
= X 2
.
In first order search methods, as the direction of searching for the maximum of the objective function F(X) the gradient vector of the objective function is selected grad(F(X k )) , to find the minimum – the antigradient vector - grad(F(X k )) . In this case, the property of the gradient vector is used to indicate the direction of the fastest change in the function:
.
The following property is also important for studying first order search methods: vector gradient grad(F(X k )) , directed normal to the function level line F(X) at the point X k .
Level lines are curves on which the function takes a constant value ( F(X) = const).
This section discusses five modifications of the gradient method:
– gradient method with a constant step,
– gradient method with step division,
– method of steepest descent,
– Davidon-Fletcher-Powell (DFP) method,
– two-level adaptive method.
In the gradient method with a constant step, the initial data is the required accuracy e, starting point of search X 0 and search step h.
X k+1 = X k – h× gradF(X k ), k=0,1,2,… (6.20)
Formula (2.58) applies if for the function F(X) you need to find the minimum. If the parametric optimization problem is posed as a maximum search problem, then to obtain new points in the gradient method with a constant step, the formula is used:
X k+1 = X k +h× gradF(X k ), k = 0, 1, 2, … (6.21)
Each of formulas (6.20), (6.21) is a vector relation including n equations. For example, taking into account X k +1 = (x 1 k +1 , x 2 k +1 ,…, x n k +1 ), X k =(x 1 k , x 2 k ,…, x n k ) :
(6.22)
or, in scalar form,
(6.23)
In general form (2.61) we can write:
(6.24)
As a rule, a combination of two conditions is used as a condition for stopping the search in all gradient methods: ç F(X k +1
) - F(X k )
ç
£
e or 
for everyone i
=1, …, n.
Consider an example of finding a minimum using the gradient method with a constant step for the same function as in the coordinate descent method:
n = 2, X = (x 1 , x 2 ), =0.1,
F(x 1 , x 2 ) = (x 1 – 1) 2 + (x 2 – 2) 2 min, h = 0,3, X 0 = (0, 1).
Let's get a point X 1 according to formula (2.45):
F(X 1 ) = (0.6–1) 2 + (1.6–2) 2 = 0.32, F(X 0 ) = (0 –1) 2 + (1–2) 2 = 2.
F(X 1 ) - F(X 0 ) =1,68 > = 0.1 continue the search.
Let's get a point X 2 according to formula (2.45):
F(X 2 ) = (0.84–1) 2 + (1.84–2) 2 = 0.05,
F(X 1 ) = (0,6 –1) 2 + (1,6–2) 2 = 0,32.
F(X 1 ) - F(X 0 ) =0,27 > = 0.1 continue the search.
Similarly we get X 3:
F(X 3 ) = (0.94–1) 2 + (1.94–2) 2 = 0.007,
F(X 3 ) = (0,84 –1) 2 + (1,84–2) 2 = 0,05.
Since the condition for stopping the search is met, found X * = X 3 = (0.94, 1.94) with accuracy = 0.1.
The search trajectory for this example is shown in Fig. 6.5.
The undoubted advantage of gradient methods is the absence of unnecessary costs for obtaining test points, which reduces the cost of one iteration. In addition, due to the use of an effective search direction (gradient vector), the number of iterations is noticeably reduced compared to the coordinate descent method.
In the gradient method, you can slightly reduce the number of iterations if you learn to avoid situations where several search steps are performed in the same direction.
In the gradient method with step division, the procedure for selecting the step size at each iteration is implemented as follows.
e, starting point of search X 0 h(usually h= 1). New points are obtained using the formula:
X k+1 = X k – h k × gradF(X k ), k=0,1,2,…, (6.25)
Where h k– step size per k th search iteration, at h k the following condition must be met:
F(X k – h k × gradF(X k )) £ F(X k ) - e× h k ×½ gradF(X k ) ½ 2 . (6.26)
If the value h k is such that inequality (2.64) is not satisfied, then the step is divided until this condition is satisfied.
Step division is performed according to the formula h k = h k ×a, where 0< a < 1.Такой подход позволяет сократить число итераций, но затраты на проведение одной итерации при этом несколько возрастают.
This makes it easy to replace and extend procedures, data and knowledge.
In the steepest descent method, at each iteration of the gradient method, the optimal step in the direction of the gradient is selected.
The input data is the required accuracy e, starting point of search X 0 .
New points are obtained using the formula:
X k+1 = X k – h k × gradF(X k ), k=0,1,2,… , (6.27)
Where h k = arg minF(X k – h k × gradF(X k )) , that is, the step is selected based on the results of one-dimensional optimization with respect to the parameter h (at 0< h < ¥).
The main idea of the steepest descent method is that at each iteration of the method, the maximum possible step size is selected in the direction of the fastest decrease of the objective function, that is, in the direction of the antigradient vector of the function F(X) at the point X k. (Fig. 2.23).
When choosing the optimal step size, it is necessary from a variety of X M = (X½ X= X k – h× gradF(X k ), h Î / h = 22(2 h-1)2=8(2h-1)=0.
Hence, h 1 = 1/2 – optimal step at the first iteration of the steepest descent method. Then
X 1 = X 0 – 1/2gradF(X 0 ),
x 1 1 =0 -1/2 = 1, x 2 1 = 1-1/2 = 2 X 1 = (1, 2).
Let's check whether the search termination conditions are met at the search point X 1 = (1, 2). The first condition is not met
F(X 1 )-F(X 0 ) = 0-2 =2 > = 0.1, but fair
that is, all partial derivatives with accuracy can be considered equal to zero, the minimum point is found: X*=X 1 =(1, 2). The search trajectory is shown in Fig. 6.7.
Thus, the steepest descent method found the minimum point of the objective function in one iteration (due to the fact that the function level lines F(x 1 , x 2 ) = (x 1 – 1) 2 + (x 2 – 2) 2 . ((x 1 – 1) 2 + (x 2 –2) 2 = const is the equation of a circle, and the antigradient vector from any point is precisely directed to the minimum point - the center of the circle).
In practice, the objective functions are much more complex, the lines also have a complex configuration, but in any case the following is true: of all the gradient methods, the steepest descent method has the smallest number of iterations, but finding the optimal step using numerical methods presents some problems, since in real problems arising When designing RES, the use of classical methods for finding an extremum is almost impossible.
For optimization problems under conditions of uncertainty (optimization of stochastic objects), in which one or more controlled parameters are random variables, a two-level adaptive search optimization method is used, which is a modification of the gradient method.
X 0
and the initial value of the search step h(usually 
). New points are obtained using the formula:
X k+1 = X k – h k+1 × gradФ(Х k), k= 0,1,2,…, (6.28)
where is the step h k +1 can be calculated using one of two formulas: h k +1 = h k + l k +1 ×a k, or h k +1 = h k × exp(l k +1 ×a k ) . The reduction factor is usually chosen l k =1/ k, Where k– iteration number of the search method.
The meaning of using the coefficient l k is that at each iteration some adjustment of the step size is made, and the larger the iteration number of the search method, the closer the next search point is to the extremum point and the more careful (smaller) the step adjustment should be in order to prevent moving away from the point extremum.
Magnitude a k determines the sign of such adjustment (with a k>0 the step increases, and at a k <0 уменьшается):
a k =sign((gradF(X k ),gradF(X))} ,
that is a k is the sign of the scalar product of vectors of gradients of the objective function at points X k And 
, Where 
=X k
–
h k ×
gradF(X k )
test point, and h k is the step that was used to obtain the point X k at the previous iteration of the method.
The sign of the scalar product of two vectors allows us to estimate the magnitude of the angle between these vectors (we denote this angle ). If 9, then the scalar product must be positive, otherwise negative. Taking into account the above, it is not difficult to understand the principle of adjusting the step size in the two-level adaptive method. If the angle between the antigradients (acute angle), then the search direction is from the point X k is chosen correctly, and the step size can be increased (Fig. 6.8).
Rice. 6.8. Selecting the search direction when
If the angle between the antigradients (obtuse angle), then the search direction is from the point X k moves us away from the minimum point X*, and the step needs to be reduced (Fig. 6.9).
Rice. 6.9. Selecting the search direction when >
The method is called two-level, since at each search iteration not one, but two points are analyzed and two antigradient vectors are constructed.
This, of course, increases the cost of carrying out one iteration, but allows for adaptation (tuning) of the step size h k +1 on the behavior of random factors.
Despite the simplicity of implementation, the steepest descent method is not recommended as a “serious” optimization procedure for solving the problem of unconstrained optimization of a function of many variables, since it works too slowly for practical use.
The reason for this is the fact that the steepest descent property is a local property, so frequent changes in the search direction are necessary, which can lead to an inefficient computational procedure.
A more accurate and efficient method for solving a parametric optimization problem can be obtained using second derivatives of the objective function (second order methods). They are based on approximation (that is, approximate replacement) of the function F(X) function j(X),
j(X) = Ф(Х 0 ) + (X - X 0 ) T × gradF(X 0 ) + ½ G(X 0 ) × (X - X 0 ) , (6.29)
Where G(X 0 ) - Hessian matrix (Hessian, matrix of second derivatives), calculated at the point X 0 :
¶ 2 F(X) ¶ 2 F(X) . . . ¶ 2 F(X)
¶ x 1 2 ¶ x 1 ¶ x 2 ¶ x 1 ¶ x n
G(X) = ¶ 2 F(X) ¶ 2 F(X) . . . ¶ 2 F(X)
¶ x 2 ¶ x 1 ¶ x 2 2 ¶ x 2 ¶ x n
¶ 2 F(X) ¶ 2 F(X) . . . ¶ 2 F(X)
¶ x n ¶ x 1 ¶ x n ¶ x 2 ¶ x n 2 .
Formula (2.67) represents the first three terms of the expansion of the function F(X) in a Taylor series in the vicinity of a point X 0 , therefore, when approximating the function F(X) function j(X) an error of no more than ½½ occurs X-X 0 ½½ 3 .
Taking into account (2.67), in Newton’s method the initial data are the required accuracy e, starting point of search X 0 and new points are obtained using the formula:
X k +1 = X k – G -1 (X k ) × gradФ(Х k), k=0,1,2,…, (6.30)
Where G -1 (X k ) – matrix inverse to the Hessian matrix, calculated at the search point X k (G(X k ) × G -1 (X k ) = I,
I = 0 1 … 0 - identity matrix.
Let's consider an example of finding a minimum for the same function as in the gradient method with a constant step and in the coordinate descent method:
n = 2, X = (x 1 , x 2 ), = 0.1,
F(x 1 , x 2 ) = (x 1 – 1) 2 + (x 2 – 2) 2 min, X 0 =(0, 1).
Let's get a point X 1 :
X 1 = X 0 – G –1 (X 0)∙grad Ф(X 0),
Where 
grad Ф(X 0) = (2∙(x 1 0 –1)), 2∙(x 1 0 –1) = (–2, –2), that is
or
x 1 1 = 0 – (1/2∙(–2) + 0∙(–2)) = 1,
x 2 1 = 1 – (0∙(–2) + 1/2∙(–2)) = 2,
X 1 = (1, 2).
Let's check whether the search termination conditions are met: the first condition is not met
F(X 1 )-F(X 0 ) = 0 - 2 = 2 > = 0.1,
but fair
that is, all partial derivatives with accuracy can be considered equal to zero, the minimum point is found: X* = X 1 = (1, 2). The search trajectory coincides with the trajectory of the steepest descent method (Fig. 2.24).
The main disadvantage of Newton's method is the cost of calculating the inverse Hessian G -1 (X k ) at each iteration of the method.
The DFT method overcomes the shortcomings of both the steepest descent method and Newton's method.
The advantage of this method is that it does not require the calculation of the inverse Hessian, and the direction – N k × gradF(X k), where N k- a positive definite symmetric matrix that is recalculated at each iteration (step of the search method) and approximates the inverse Hessian G -1 (X k ) (N k ® G -1 (X k ) with increase k).
In addition, the DFT method, when used to find the extremum of a function of n variables, converges (that is, gives a solution) in no more than n iterations.
The computational procedure of the DFT method includes the following steps.
The input data is the required accuracy e, the starting point of the search X 0 and initial matrix N 0 (usually the identity matrix, N 0 = I).
On k th iteration of the method, the search point X k and the matrix are known N k (k = 0,1,…).
Let us denote the search direction
d k = -N k × gradФ(Х k).
Finding the optimal step size l k in the direction d k using one-dimensional optimization methods (in the same way as in the steepest descent method, a value was selected in the direction of the antigradient vector)
H. Let us denote v k = l k × d k and get a new search point X k +1 = X k + v k .
4. Check that the condition for stopping the search is met.
If ½ v k ½£ e or ½ gradF(X k +1 ) ½£ e, then the solution is found X * = X k +1 . Otherwise, we continue the calculations.
5. Let us denote u k = gradФ(Х k +1) - gradФ(Х k) and matrix N k +1 calculate using the formula:
H k +1 = H k +A k + B k , (6.31)
Where A k =v k . v k T /(v k T × u k ) , B k = - H k × u k . u k T . H k /(u k T × H k × u k ) .
A k And IN k are auxiliary matrices of size n X n (v k T corresponds to a row vector, v k means column vector, the result of multiplication n-dimensional line on n-dimensional column is a scalar quantity (a number), and multiplying a column by a row gives a matrix of size n x n).
6. Increase the iteration number by one and go to point 2 of this algorithm.
The DFT method is a powerful optimization procedure that is effective in optimizing most functions. For one-dimensional optimization of the step size in the DFT method, interpolation methods are used.
The term search engine optimization comes from the English “search engines optimization” - “optimization for search engines”. This direction has been increasingly developed in the RuNet in recent years and represents one of the main and most effective methods responsible for the success of any Internet project.
Effect– stable high positions in search engines.
Correcting errors in site navigation and editing program code is work on the internal factors of the site that affect both the convenience of the site for users and its “friendliness” in relation to search engine robots.
Content growth– adding new pages containing information useful to target visitors.
Placing links on thematic resources differs from a chaotic exchange of links in that links are published only on sites whose visitors may be truly interested in the information posted on the pages of your site.
The use of “white hat” optimization, as a rule, leads not only to an increase in the site in the first positions, but also to an increase in site visitors by several tens of times.
What problems does search engine optimization solve?
Search engines are the most important tool for navigating the Internet today. With their help, they search for information on the Internet, compare, analyze, ask for advice, look for like-minded people, acquaintances, and even the meaning of life. If previously catalogs were a more popular tool for navigating the Internet, today their volume and ramifications, growing along with the increase in volumes of information, have grown so much that they have either become overly complex for the user or contain very little information. At the same time, the quality of search engines has improved significantly over the past few years. Therefore, it is not surprising that users switched en masse to search engines.
By becoming the most popular sites on the Internet, search engines have gained an additional size effect. Now these are not only the most visited sites, but also the most famous. As a result, when a user goes online for the first time, he first goes to the site that he already knows from friends, from the press or from offline advertising, that is, to a search engine.
This state of affairs will continue for a long time, since a significant proportion of users are not very familiar with computers in general, and the number of such users in our country is apparently growing. Not too savvy Internet users use the search bar of the machine as the navigation bar of the browser. Many users do not distinguish between the concepts of “Internet” and “search engine” at all. This can be clearly seen by the number of search queries containing the site address.
That’s why this is so important for many companies optimization site, how the process of achieving first places in search results in search engines for company-targeted queries. Why is it important to take one of the first places? Because users, on average, view one page of search results and rarely go to sites that only have links on the second page and beyond.
Why is there a second page! A study of user transitions from Yandex search results, conducted several years ago by SpyLOG and Ashmanov and Partners, showed that the share of transitions that can be expected on the seventh line tends to zero, that is, sites that are below sixth place in the search results , – also “overboard”. Above the search results there are advertising lines, which are increasing in number over time. They also reduce the number of clicks that sites get in search results, because for users these are also search results.
By the way, for North American sites the share of transitions from search engines from all transitions is on average 60%, and for corporate sites it is even higher. That is, more than half of all visitors receive the average website from a search engine. On the Russian-language Internet, the share of search traffic is lower, but still very large and constantly growing.
That is why optimization today is a large, branched market of services, in which there are significantly more players than in the online advertising market. And the volume of this service market in Russia in 2008 is estimated at $200 million, which is only three times less than the advertising market. How could it be otherwise, if the effectiveness of this marketing method is in no way lower than other advertising tools on the Internet!
Optimization is a set of techniques and methods that allow you to reach the top lines in search results. All optimization techniques are open and described on numerous forums, specialized optimizer sites and in countless articles. It is very important that there are no “secret” optimization methods. Everything here is transparent and has long been known. The greatest importance in the speed of obtaining optimization results is the experience of the optimizer, that is, the ability to quickly assess the situation and choose the right methods of work, however, even a beginner, armed with patience and perseverance, can achieve excellent results.
WHEN STARTING THE OPTIMIZATION PROCESS, BE CLEARLY UNDERSTANDING THE FOLLOWING POINTS.
1. This is a process of trial and error. Despite the fact that there are fairly accurate “recipes for success,” in each specific case they may not work, and the likelihood of their working is lower, the more fellow optimizers are working on the same search words and in the same market sector. It is necessary to try all optimization methods to achieve results.
2. Optimization is a long process. Even if you quickly make changes to the site, the search robot will not update information about the site in the database immediately, but at best in a few days, or rather even in a week. That's why the optimization process usually drags on for many months and all the results come very gradually.
3. Optimization is a very painstaking process, where many factors must be taken into account: the features of each search engine, the features of the market in which the company operates, the activity of competitors and the actions they take, and so on. Moreover, all these factors must be taken into account constantly, and not just once at launch.
4. Optimization is an unstable process. Search engine algorithms are constantly changing, and the market landscape is also changing due to competitors and the optimization efforts they take. Therefore, the successes that the company achieved a few days (weeks) ago may turn into nothing today. Hence, optimization must be done constantly.
5. Search engines resist SEO efforts because they degrade search quality. Search engines strictly and unilaterally regulate acceptable behavior of optimizers and unceremoniously remove from search results (search results) sites that, in their opinion, do not comply with these rules. Moreover, these rules are not published and are constantly changing, so any optimization action tomorrow may be “outside the law.”
Why do search engines fight SEOs?
As a result of the actions of optimizers, search results change. Random results disappear from the search or go downhill - links to forums, links to pages that have long disappeared. This undoubtedly makes the search better: its accuracy increases. However, at the same time, the completeness in the results, that is, the search engine’s coverage of various topics related to the search query, sharply decreases. For example, the query “car” includes a whole range of different interests: buying a new or used car, rental, repair, device, spare parts, history, abstract, types, etc., etc. At the same time, search engines are all as one they issue either sale (new, used) or rental cars. In rare cases, we also come across the sale of spare parts. Thus, more than half of the possible interests of users were dropped from the search results (on the first few pages), meaning many users do not receive the information they need and will be forced to clarify them repeatedly. Compare the search results for the same word on the Yandex search engine (Figure 5.9) or Google with the search results of Nigma (Figure 5.8) - a machine that clusters search results by different topics - you will see how few different topics come to the top search pages of the “big” search engines.
Many SEO specialists talk about all sorts of magical ways to “bring your site to the TOP 3 in two weeks.” Any thoughtful manager has long known the principle of free cheese and understands that quality work always costs a lot. The main principle of search engine optimization is simple - be more unique, faster and more authoritative than your competitors. In the hands of a skilled SEO professional, SEO tools can cause an increase in traffic and conversion, while the work of a self-taught SEO can end in a site ban and disaster for your campaign.
Internal website optimization
The structure of the resource is one of the most important elements of promotion. If page tags are not aligned with popular queries and the link map is poorly designed, neither link buying nor increased behavior will help the site get to the top. Internal optimization is the painstaking work of analyzing and finalizing materials to the required line - between user readability and saturation with important key phrases. You can read more about the quality of texts in this article.
Link building
The citation and authority rating of a resource is the number of links left by users on external sources. In theory, an organic link is a recommendation for a search engine, a sign that the page being linked to is of high quality and trustworthy. Naturally, this approach to ranking the authority of sites has given rise to a huge market for link traders and a huge number of resources, the purpose of which is only to increase the link mass for money, to artificially increase it. Google and Yandex search engines fight such link brokers by excluding them from search results and reducing the importance of their links to zero. In 2013, it makes sense to purchase only high-quality, permanent links from authoritative resources (long domain name age, high PR, presence in large directories and huge traffic). Such links are expensive, however, they bring additional traffic and practically guarantee a high ranking for your site.
Emulation of behavioral factors
The most important trend of the new year is that search begins to focus on people. Many optimization specialists have been successfully using behavioral emulation for a long time. For example, if a user enters the query “The most delicious coffee in the city” and finds the page “Coffee shop 1”, goes to it, looks at several pages for 3-10 minutes, and then goes on to add the page to his favorites - this is a good sign for the search engine. that "Coffee House 1" is really good. Let's look at another example. For the same request, there is a position “Coffee shop 2”, from which users run away after 15 seconds and never return. This is a huge minus, which proves that “Coffee Shop 2” is dull and uninteresting. Skillfully using this knowledge, a competent SEO specialist can quickly and without huge investments raise a resource to the TOP10 or even TOP3.
Social Marketing
A separate topic is website optimization for the needs of social networks (so-called SEO 2.0). The main idea of these measures is to write and post information on social networks that encourages discussion and, as a result, an increase in traffic. Social marketing also includes user likes (Google+, Facebook). Clicks by your visitors on the buttons of these networks can and will increase your ranking in search results.
Lecture 5. Information retrieval optimization: recommendations
Information search processusually of an empirical nature. In this case, the user’s behavior, as the organizing beginning of managing the search process, is motivated by:
- information need,
- the variety of strategies, technologies and tools provided by the system.
To evaluate adequacy of query expressionand the completeness of the result obtained, the user can:
- find additional information,
- organize the search process so that part of the search results can be used to confirm or deny the adequacy of another part.
Operational objects directly involved in user interaction with the search engine aredocument search image And search query image.
Search document image - description of a document, expressed by means of information retrieval language and characterizing the main semantic content or any other features of this document necessary for searching it on request.
Information retrieval technologies- p Search tools and technologies used to fulfill information needs. They are determined by the type of problem being solved by the librarian: the ratio of his knowledge and ignorance about the object under study. In addition, the process of interaction between the librarian and the system is determined by the level of his knowledge of the functionality.system as a tool and knowledge of the content of the resource (completeness of information presentation, reliability of the source, etc.).
The procedure for finding the necessary information must be divided into four main stages:
1. Organization of the search
- definition of knowledge area;
- selection of data type and sources;
2. Implementation of the search process
- organizing a search based on the initial fragments of a word,
- the wording of the “search image”,
- final refinement search
It is generally accepted to organize a search based on the initial fragments of a word (search with truncation on the right). For example, instead of the word “library” you can enter its fragment « library*". In this case, documents will be found that contain not only the word « library", but also"library", "librarian", " library science”, etc. In each individual case, the librarian must clearly understand what needs to be found (“search image formulation”), since in the option proposed to him a much larger number of documents will be found than when specifying the word in full (without truncation). In the resulting array of information, you can conduct a clarifying search and, as a result, obtain more relevant and relevant data.
3. Search results evaluation
The search result must meet the following requirements:
- uniqueness,
- fullness,
- consistency
Different types of search determine different requirements for the functionality of the search system in terms of evaluating the result. However, for the casesubject searchthe proof of completeness is a priori: the search result itself confirms the existence (or absence) of an object that has the desired properties. In this case the resultthematic searchis multiple and requires subsequent systematization - another procedural step for ordering the resulting set of objects according to the values of a basis that is not explicitly defined. In turn,problematic searchalready presupposes a two-level systematization.
Complication, specification and detail of the request, in case of an unsatisfactory answer, are carried out by modifying the search image, that isreformulationrequest and conduct a second search in the same data array that was obtained as a result of the initial search. It is advisable to double-check the same request in different search engines (for example, a combination of Google and Yandex).
4. Processing search results
- selection of the most useful information;
- choice of information processing method
- search for patterns, formal rules and structural connections in the collected information;
- creative interpretation of the results obtained;
- integration of the extracted “knowledge”.
The development of the search process is carried out by modifying the POS expression, by reformulating the query and conducting a second search in the same data array or in a subarray obtained as a result of the initial search.
Methods for processing search results
Based on the nature of the transformations (in the context of further use of processing results), methods for processing search results can be divided into two groups:
- Structural and format transformations;
- Information-analytical (logical-semantic, structural-semantic transformations).
Search engines are used directly for searching, the number of which amounts to several hundred in the world. They focus on certain types of queries or a combination of them (bibliographic, address, factual, thematic, etc.). In addition, there are full-text, mixed and other search engines.
References
