 In the current situation, property prices are a well-known phenomenon. According to the area where the house is located, decoration, and construction - the price is always different. And this list of features really does affect the price. Finally, when a client's budget and expectations match the price of the home, a transaction occurs. But considering all these variable characteristics, it is difficult to predict prices. Machine learning can help solve this problem effectively. Today we will try to establish a real estate price forecast model. The model is a well-defined programmable agent whose primary purpose is to improve its performance based on available resources. In our case, we expect to improve house price performance and the characteristics are: crime rate, area, industrial area, nitric oxide concentration, average number of rooms per dwelling, age, distance to the main 5 sites, to the road The distance, the tax, the student-teacher ratio and the lower status of the population. If we represent all this in the table, then it is so, we can see that all eigenvalues are converted into values. In addition, our estimated / predicted house price is also a value, as the price can be \$ x, \$ x can be any decimal value. For the sake of simplicity, we treat all resources as and. According to the y function, the price \$ x will be different. Simply put, we want to know how big the change in y is and x will be different. We want to find the relationship between them. In this particular case, this relationship is our model, which uses the y resource and maps the output to the x price.

## Software Who help in Prediction

Let us now try to develop an equation so that it can give our predictions in layman's terms. We definitely know that \$ x is not equal to y. This is like saying that the crime rate is not the price of the house, which is obviously correct. But x and y are related. The crime rate will affect housing prices, which is practical. If we want to present this mathematically, we can have alpha * \$ x = y so. We know that the crime rate will affect the price of the house, but what affects the price of the house is the alpha. There is a big problem now. We do not depend on alpha. In fact, this is a default value with other known values in each row entry in the table. Just as the crime rate is 0.0689%, the price is \$ 200 and then in the simplest form, alpha is like 200 / 0.0689 (for example, this value is m). Simple? Now we have the alpha value. So far, what we have done is the training phase of the well-known simple model. This model is known as linear regression and is expressed in the simplest terms. Now the model is ready for prediction, the equation is as follows, y / m = \$ x. Now we know that alpha is m. For each new entry in the future, we need to know the crime rate. For this purpose, the price of the house will be the crime rate, that is, y multiplied by alpha, which is m. Now, from this model, we can get the value of \$ xo the price we are looking for. Simple, is not it? In fact, this is not so simple. For each resource, such as crime rate, the process continues Finally, many resource variables appear, such as y1, y2, y3 ... Finally, all the equations are combined into one. Among them, there is a loss function that determines the loss. Finally, we obtain coefficients such as alpha to map future values.

• Share On