Penyelidikan ini mencadangkan pembaikan pada rangka kerja ditemui melalui pemahamn untuk menganalisis dan peramalan data siri masa dengan pola musim dan trend. Model Peramalan boleh dikategorikan sebagai kualitatif atau kuantitatif. Rangka yang dibangunkan dalam kajian ini bertujuan untuk model kuantitatif.
Secara total, lapan model kuantitatif, termasuk di dalam rangka, iaitu Additive Decomposition Method, Multiplicative Decomposition Method, No trend, linear trend dan quadratic trend Simple Exponential Smoothing, Holt’s Trend Corrected (Double) exponential Smoothing, additive dan multiplicative Holt’s – Winter (Triple) Exponential Smoothing.
Penyelidikan dilaksanakan dalam tiga tahap. Pada tahap pertama, sebuah rangka untuk mengenal pasti bahagian-bahagian data siri masa dan untuk memilih model peramalan yang sesuai daripada lapan model yang dinyatakan di atas untuk dibangunkan. Rangka ini dipaparkan dalam bentuk diagram alur.
Pada tahap kedua, perisian berasaskan Microsoft Excel dibangunkan untuk melakukan langkah-langkah untuk kelapan-lapan model. Dari perisian, grafik data sebenar melawan data dijangka serta anggaran kesalahan untuk model tersebut dihasilkan. Anggaran kesalahan adalah Mean Absolute Deviation (MAD), Mean Squared Error (MSE) dan Mean Absolute Percentage Error (MAPE). Pada tahap akhir kajian, rangka dan perisian yang diterapkan pada lapan set data yang diterbitkan dalam buku dan internet. Keputusan yang diperolehi untuk aktiviti ini menunjukkan bahawa rangka dan perisian boleh digunakan untuk analisis siri masa dan peramalan untuk kes data bermusim dan trend.
___________________________________________________________________________________
This research proposes improvements to the frameworks encountered in the literature for analyzing and forecasting time series data with seasonality and trend. Forecasting models can be categorized as qualitative or quantitative. The framework developed in this research is intended for quantitative models. In total, eight quantitative models were included in the framework, namely the Additive Decomposition Method, Multiplicative Decomposition Method, No trend, linear trend and quadratic trend Simple Exponential Smoothing, Holt’s Trend Corrected (Double) exponential Smoothing, additive and multiplicative Holt’s – Winter (Triple) Exponential Smoothing.
The research was carried out in three stages. In the first stage, a framework for identifying the components of the time series data and for selecting the appropriate forecasting model out of the eight models listed above was developed. The framework was expressed in the form of a flowchart.
In the second stage, Excel-based software were developed in order to perform the steps for each of the eight models. From the software, graphs of the actual data versus the forecasted data as well as error estimates for the model were generated. The error estimates are the Mean Absolute Deviation (MAD), Mean Squared Error (MSE) and Mean Absolute Percentage Error (MAPE). In the final stage of the research, the framework and software were applied to eight data sets that were published in books and the Internet. The result obtained for this exercise shows that the framework and software can be used for time series analysis and forecasting for the case of seasonal and trend.
Framework for time series analysis and forecasting_Muhammad Farid Mohd Hassan_E3_2011_NI