How to confuse analytics - 4. Probability and accuracy
In the last article, I said that numeric attributes are directly related to the operations that we perform on objects. Moreover, the natural numbers are the simplest of the attributes we are considering. There are more complex ones. For example, matrices. If we are talking about the property of a linear transformation in three-dimensional space, then it is written with 9 numeric values, from which it is convenient to form a 3 by 3 matrix. The reason for this is that two transformations performed sequentially are also transformations, whose numerical attributes can be obtained by multiplying two matrices. This is the power of modeling the transformation using the matrix.
I would give a lot to teach mathematics in this way: through a practical task, through the input of the necessary objects (numbers, matrices, wave functions) and an explanation of how operations on them help solve specific problems. That was how the training in the Physics and Mathematics School was built, in which I had the opportunity to study - in boarding school No. 18 at Moscow State University, thanks to the teachers!
So, we have created a repository where you can put information about nouns, adjectives and verbs.
We have divided flies and cutlets - modeling of objects of accounting from their interpretation. It all came down to the fact that there are objects of accounting (4-dimensional spatial volumes), the mereological relations between them and the interpretation of these relations through classification. The only thing that is new to us is a way of thinking about time as another dimension. For example, on the time axis, you can make marks that allow you to bind to specific moments and measure the time intervals modeled by the attributes "with:" and "by:".
')
Imagine being asked to remember when was your last X-ray? Most likely, you will remember the year, maybe a month. I have a friend who remembers the day, hour and name of the doctor, because she has a phenomenal memory. But I can only say about a month. Now imagine that you are at the doctor who asks you about fluorography, and he has an information system on which he must enter the date of your last X-ray examination. Let's say that you remember that it was in May. The doctor, most likely, will say: let us write for definiteness: on May 15th. Nothing like? You may have encountered these kinds of situations.
When the IP designer indicates the way data is stored about the accounting object, it indicates the type of data. If we are talking about time, then the methods of storing information about time and time intervals are standardized. You can record the time to within a fraction of a second. And this, at times, matters. However, in the example with fluorography, I have to say this: May last year plus or minus a month. Have you ever seen the format for recording time data so that you can record this information that way? I saw this opportunity in BusinessStudio. There you can say that the time of passage of fluorography - June 15 with a dispersion of one month and the Gaussian form of distribution (which, by the way?).
Another example: let you create a work plan for the next year. And let you know that work on the replacement of the ground wire will begin in the first half of May. How to record this information in the IP? Perhaps you will create a separate work plan for the month of May, and indicate that in May you need to start changing the ground wire. This record means literally the following: the beginning of the work on the replacement of the ground wire is determined by a uniform rectangular distribution a month wide.
This suggests that we model both the past and the future using distributions. If we talk about the past, we model the accuracy distribution of our knowledge about it. If we talk about the future, then we model the probability distribution of certain events. In either case, we model the allowable intervals in which the past or future lies.
For example, speaking of the future, we can say that someone from the sales department will be the executor of the operation for concluding a contract, but we do not know exactly who it is. If we talk about the past, then we can say that someone from the sales department was the executor of the sale, but we do not know who exactly. Both thesis and the other are modeled in the same way - using distributions.
So, the question of what we are modeling can be answered as follows: we model objects of accounting, classify them and do it with some probability if it is about the future, or with some accuracy if it is about the past.
It may seem that the mention of accuracy and probability is unnecessary, but in practice the consideration of these nuances has a significant impact on the decision. Example:
Suppose there are three events measured with the same accuracy - plus or minus one hour. Let the first event happen somewhere from 16-00 to 18-00, the second - somewhere from 17-00 to 19-00, the third - somewhere from 18-00 to 20-00. And let the task is to build events in the order of their accomplishment. It is clear that there are 3 options: with probability one fourth: 2, 1, 3, with probability one second: 1, 2, 3 and with probability one fourth - 1, 3, 2.
It is clear that decision making becomes probabilistic. On the first course of the Physics and Technology Institute, the first lab is the most difficult to pass; it is dedicated to the accuracy of measurements and the probability of our ideas. It comes to students very tightly that school physics is lying to us. In reality, there are no exact answers and solutions.
I would give a lot to teach mathematics in this way: through a practical task, through the input of the necessary objects (numbers, matrices, wave functions) and an explanation of how operations on them help solve specific problems. That was how the training in the Physics and Mathematics School was built, in which I had the opportunity to study - in boarding school No. 18 at Moscow State University, thanks to the teachers!
So, we have created a repository where you can put information about nouns, adjectives and verbs.
We have divided flies and cutlets - modeling of objects of accounting from their interpretation. It all came down to the fact that there are objects of accounting (4-dimensional spatial volumes), the mereological relations between them and the interpretation of these relations through classification. The only thing that is new to us is a way of thinking about time as another dimension. For example, on the time axis, you can make marks that allow you to bind to specific moments and measure the time intervals modeled by the attributes "with:" and "by:".
')
Probability and accuracy
Imagine being asked to remember when was your last X-ray? Most likely, you will remember the year, maybe a month. I have a friend who remembers the day, hour and name of the doctor, because she has a phenomenal memory. But I can only say about a month. Now imagine that you are at the doctor who asks you about fluorography, and he has an information system on which he must enter the date of your last X-ray examination. Let's say that you remember that it was in May. The doctor, most likely, will say: let us write for definiteness: on May 15th. Nothing like? You may have encountered these kinds of situations.
When the IP designer indicates the way data is stored about the accounting object, it indicates the type of data. If we are talking about time, then the methods of storing information about time and time intervals are standardized. You can record the time to within a fraction of a second. And this, at times, matters. However, in the example with fluorography, I have to say this: May last year plus or minus a month. Have you ever seen the format for recording time data so that you can record this information that way? I saw this opportunity in BusinessStudio. There you can say that the time of passage of fluorography - June 15 with a dispersion of one month and the Gaussian form of distribution (which, by the way?).
Another example: let you create a work plan for the next year. And let you know that work on the replacement of the ground wire will begin in the first half of May. How to record this information in the IP? Perhaps you will create a separate work plan for the month of May, and indicate that in May you need to start changing the ground wire. This record means literally the following: the beginning of the work on the replacement of the ground wire is determined by a uniform rectangular distribution a month wide.
This suggests that we model both the past and the future using distributions. If we talk about the past, we model the accuracy distribution of our knowledge about it. If we talk about the future, then we model the probability distribution of certain events. In either case, we model the allowable intervals in which the past or future lies.
For example, speaking of the future, we can say that someone from the sales department will be the executor of the operation for concluding a contract, but we do not know exactly who it is. If we talk about the past, then we can say that someone from the sales department was the executor of the sale, but we do not know who exactly. Both thesis and the other are modeled in the same way - using distributions.
So, the question of what we are modeling can be answered as follows: we model objects of accounting, classify them and do it with some probability if it is about the future, or with some accuracy if it is about the past.
It may seem that the mention of accuracy and probability is unnecessary, but in practice the consideration of these nuances has a significant impact on the decision. Example:
Suppose there are three events measured with the same accuracy - plus or minus one hour. Let the first event happen somewhere from 16-00 to 18-00, the second - somewhere from 17-00 to 19-00, the third - somewhere from 18-00 to 20-00. And let the task is to build events in the order of their accomplishment. It is clear that there are 3 options: with probability one fourth: 2, 1, 3, with probability one second: 1, 2, 3 and with probability one fourth - 1, 3, 2.
It is clear that decision making becomes probabilistic. On the first course of the Physics and Technology Institute, the first lab is the most difficult to pass; it is dedicated to the accuracy of measurements and the probability of our ideas. It comes to students very tightly that school physics is lying to us. In reality, there are no exact answers and solutions.
Source: https://habr.com/ru/post/332668/
All Articles