DonNTU   Masters' portal   FCST   Department ACS

Perevozchikov Alexander

Faculty of computer science and technology (CST)

Department of Automated Control Systems (ACS)

Speciality Information Control Systems and Technologies

Computerized subsystem of forecasting and planning of procurement of raw materials

Scientific adviser: Ph.D., Alexander Sekirin

Resume Abstract

Abstract

Contents

• Introduction
• 1. Relevance of the theme
• 2. Review of existing systems
• 3. Description of the object of research
• 4. Goals and objectives of the master's work
• 5. Expected scientific results
• 6. Review of the methods of forecasting
• 7. The mathematical formulation of the problem of forecasting
• Conclusion
• Source list

Introduction

In the present conditions of economic independence for many industrial companies have become very important issue on the application of forecasting in its activities. In drawing up the plan of production there are very important two questions: business opportunities of the company and the demand for its products. Today, when businesses are forced to work on the «laws of the market», employers want to know the prospects of the enterprise, to look into the future to assess the possible ways of development, to predict the consequences of certain decisions. Forecasting and planning are a powerful tool to not only effectively harness the power of the enterprise, but also to minimize the risks associated with the purchase of raw materials. In modern conditions is practically impossible to meet the company which does not apply forecasting and planning for the production. This is due to the fact that in the course of its work the employer can face two situations:

1. there is no demand for the products, and it is in stock, without making profit;
2. products out of stock due to high demand.

Any of these situations detrimental to the profitability, given the growing competition in today's market,   the use of procurement logistics becomes mandatory for the effective operation of the business.

1. Relevance of the topic

Every job requires a person thinking about the goals of the order and the possible results. This coordination of the planned measures, mediated plan helps to achieve success at a lower cost for the «minimum means – the maximum result». Economic projects and transactions aimed at achieving the objectives (profit, growth or profitability the conquest of the market) require prior understanding as a forecast, plan or program of action as an indispensable conditions justify the design and reliability of obtaining the desired result.Effective activities of enterprises and firms in market economy is largely dependent on how significantly they foresee the far and the near prospect of the development, that is, from forecasting.In the present conditions of economic independence for many industrial companies have become very important issue on the application of forecasting in its activities. In drawing up the plan of production there are very important two questions: business opportunities of the company and the demand for its products. Today, when businesses are forced to work on the «laws of the market», employers want to know the prospects of the enterprise, to look into the future to assess the possible ways of development, to predict the consequences of certain decisions. Thus forecasting and planning are an integral part of any modern production, is being implemented all over the place and, in fact, is the criterion to maintain competitiveness in the market.

2. Review of existing systems

The analysis considered the following system for forecasting and planning:

1. 1С: Управление торговлей 8 [6];
2. Regression Analysis and Forecasting [7];
3. GeneXproTools [8];
4. CatMV [9].

System 1С: Управление торговлей 8 enables us to meet the challenges of monitoring the date of delivery of the goods, the cost of funds and timely payment to suppliers.Programma allows you to register differences in the reception of goods, to analyze the cause of failure of supply (failure of suppliers from the supply of goods) account for additional services and additional costs the supply of goods.

Regression Analysis and Forecasting uses a range of commonly used statistical measures to test the reality of the analysis, the results calculating in text for ease of use. Once the relationship has been defined, prediction can be made based on a wide range of methodologies used. Clear step by step use allows for projects to develop forecasts the temporary manner. The main feature of multiple analysis and forecasting model is the simplicity and flexibility of input, with attached certificate requests; user–friendly display of the results for non statistics, and multiple the only independent variable in the regression, and fast forecasting process with options to use the three polynomials two polynomials, exponential or linear trend lines on independent variables.

GeneXproTools can handle data sets with tens of thousands of variables and retrieve the most desired features and their relationship. It is also very user friendly program that makes it easy to access all types of data stored from raw text files to databases and Excel spreadsheets. If you want to integrate the generated model with other applications, the program will allow you to translate them into sixteen different languages, from major languages such as: Java and c#, to specialized languages such as: Matlab and VHDL.

Programm CatMV is the realization of «Caterpillar»–SSA method of analysis of time series, which may contain missing values. Applicable findings will lead to the recovery of additional components of the time series such as trends and periodic components, with simultaneous filling of the missing data. This program allows you to perform prediction to add the missing values after a certain point of time series.

After a detailed analysis of each of these systems was decided against their use, as they were identified the following common faults:

1. not explain the obtained solutions (do not know what kind of forecasting methods used);
2. often do not perform all the necessary functions;
3. have a relatively high cost.

3. Description of the research object

The object of research is the process of procurement management in the food industry. Procurement management on food business is a particularly important issue because properly drawn up a plan and forecast on the one hand to minimize the volume of stored goods, leading to a reduction in storage costs and, consequently, improved economic efficiency of the enterprise, and on the other – can prevent damage to the product itself (as the food products require special storage conditions).

4. Goals and objectives of the master's work

The aim is to develop a computerized system that allows forecasting of demand and, on the basis of forecasting, planning, procurement of raw materials at the plant in order to optimize costs, increase economic efficiency enterprise and, consequently, increase profits. To achieve this goal we must solve next tasks:

1. domain analysis and existing methods, models and algorithms for planning and forecasting;
2. development of components and generalized models of forecasting procurement;
3. determining the accuracy and the selection of the effective parameters of the model prediction;
4. develop a model of procurement planning, using data prediction;
5. develop recommendations based on the models of forecasting and planning to improve the efficiency of the enterprise.

5. Expected scientific results.

As a result, the development of a computerized subsystem of forecasting and planning of procurement of raw materials   planned to achieve the following research results:

1. forecasting models to forecast purchases will allow us to achieve accuracy to at least 85%;
2. planning model developed will allow us to achieve to plan purchases for up to 1 year, with an accuracy at least 85%.

6. Review of the methods of forecasting

At the present stage of development of the market economy, there are many methods, models, algorithms and packages ready–made application allows you to plan and forecast with the required accuracy. Let us consider the main ones. Fuzzy logic. Most methods do not allow prediction function with quality indicators, while theory of fuzzy logic provides a convenient tool for supplying expert of the rules of the market in mathematical form, provides automatic establishment of the model parameters with regard to quantitative and qualitative indices [2]. The main disadvantages of fuzzy logic are:

1. lack of a standard methodology for constructing fuzzy systems;
2. impossibility of mathematical analysis of fuzzy systems by existing methods.

Regression analysis. It is used mainly in the medium–term forecasting, as well as in long–term forecasting. Medium–and long–term given the opportunity to establish changes in the business environment and account for the effects of these changes on the monitoring indicator [1]. The simplest version of the regression model is a linear regression. This is most commonly used in practice because of its simplicity. The main disadvantage of non–linear regression models is the complexity of determine the type of functional dependence, as well as the complexity of the definition of the model parameters. The disadvantages may also include the fact that, in practice, most of the dependencies are non–linear prediction, which makes use this method impractical.

Genetic algorithms (genetic algorithm, GA). It is often used to solve optimization problems, as well as search problems. However, some modifications GA can solve the problem of forecasting [3]. Prediction algorithm based on GA can take more than 15 external factors using the base GA.Main advantages of genetic algorithms:

1. can be used for a wide class of problems;
2. simple and transparent in the implementation;
3. can be used in problems with the changing environment.

The disadvantages of genetic algorithms include the following:

1. not guarantee finding a global solution in a reasonable time;
2. not guarantee that the solution found is the optimal solution;
3. in cases where the problem can be solved by specifically designed for this method, it is almost always will be more efficient, both in speed and accuracy obtained solutions.

Neural Networks. Given the shortcomings of previous models, the most suitable for the prediction of purchase will be neural networks [5]. There are many applications of neural networks to solve the problem of time series prediction. Usually when forecasting time series are used multilayer, often three–layer, feedforward neural networks. If we want a way out, the network has a manner, the structure shown in Fig. 1.

The main advantages of neural networks are:

1. nonlinearity. Neural networks allow for non–linear dependence of the output signal from the input;
2. adaptability. Neural networks have the ability to adapt their synaptic weights to changes environment. Moreover, for operation in nonstationary environments (where the statistics change with time) can be created neural network synaptic weights changing in real time;
3. fault tolerance. Neural networks are implemented on the basis of electronic components, potentially fault–tolerant, as contextual information is distributed to all the neural network connections.

Major drawback of neural networks is that the developer is not available what is happening inside the network [10]. It forming inputs, then calculates and outputs simply compares with one another. There is no possibility in detail and step trace is obtained as the output values have been calculated. This mode of performing calculations in a «black box» extremely complicates the interpretation of the results and modify the network – it is not clear that it must change in order to become skillful.

7. The mathematical formulation of the problem of forecasting.

Prediction procurement of raw materials is an integral part of the functioning of the food industry. This is a complex, multifactorial process that involves market analysis, and analysis of the company and the organization production and sale of goods [4]. The main purpose of forecasting – the definition of market trends and the company, in a constantly changing factors, and a plan based on the prediction obtained, which would be allowed to increase performance of the enterprise. Thus the problem reduces to finding the optimal quantitative values of goods to be procured based on the values of the respective factors. In general, it looks like this:

                                                     yⁱ_j+1 = f(x ⁱ_j1 , x ⁱ_j2 ,x ⁱ_j3 ... x ⁱ_jn ,y ⁱ_j ),

where
• y – the volume of production,
• x – factors affecting the volume of production;
• i – number of products;
• j – number of the period of time for which the forecast is based;
• j+1 – number of the period of time for which the forecast.

Generally target value y can be represented as a state variable or output variable of a management system, the input of which is fed the appropriate input and control quantities u and which is under the influence the respective interference s (Fig. 2).

Resume