p–ISSN: 2723 - 6609 e-ISSN: 2745-5254
Vol. 5, No. 01 January 2024 http://jist.publikasiindonesia.id/

Doi: 10.59141/jist.v5i01.806 309

DASHBOARD USING R SHINY FOR ARIMA AND PROPHET COMPARISON

Christopher Caleb
Universitas Tarumanagara Jakarta, Indonesia

*Correspondence
ABSTRACT

Keywords: Solusi Dokumen,
Forecasting, ARIMA, Prophet.

As technology progresses, more and more business processes are
digitised. This research was made to determine how the document
industry is affected and make an educated guess on revenue forecasts
based on previous data. Data will be taken daily for five years, from
January 2017 to December 2022. Revenue data will be forecasted using
two methods: ARIMA and Prophet. Data will be plotted on a graph in a
user-authenticated dashboard made with R Shiny. The results of each
forecast will be error-tested with MAE and RMSE. The results show that
Prophet consistently produces a smaller number in both tests, which
shows that Prophet is the more accurate method.

Introduction

A significant boost in office digitalisation, mainly documentation, occurred with
the COVID-19 virus pandemic, which WHO declared as a public health emergency in
January 2020 the impact of COVID-19 on the workforce at various levels (Ella & Andari,
2022). Industry 4.0 ideas, solutions, and digital transformation have become a panacea
for many affected by the pandemic. By moving towards digitisation, companies are
reducing the need for paper and providing access to information to all employees without
the costs associated with printing and storing physical paper documents (Jamaludin et al.,
2022).

This research is a case study of a company that provides document solutions.
Document solutions provide transformation products and services, from hardware-based
service providers to solution-based services that cover all aspects of the document
lifecycle, from input (creating, scanning, merging, editing, capturing), digital document
management (sharing, indexing, storing, archiving, distributing), to document output
(printing, faxing, scanning, copying, emailing, web-viewing) (Aksenta et al., 2023).

In all areas of business, planning and decision-making are necessary. For effective
implementation, it is necessary to carry out proper analysis. One of the fruits of a mature
analysis is forecasting with a solid foundation (Sahrudin, 2016). Forecasting covers many
fields, including business and industry, government, economics, environmental,
medicine, social sciences, politics, and finance. Predictions can be made for the next few
years or even just a few minutes. Some things are more accessible to predict than others.
The accuracy of predictions depends on several factors, including:
1. how deep the understanding is related to the factors involved;
2. how much data there is;
3. and whether the prediction targets will be affected by making predictions.

Christopher Caleb

Jurnal Indonesia Sosial Teknologi, Vol. 5, No. 01, January 2024 310

Most prediction cases involve the use of time-series data. A time series is a
chronological sequence of observations of variables (MANOGARI, 2019). Many
business forecasting applications use daily, weekly, monthly, and yearly data but are not
limited to those time vulnerabilities (Aisah, Zaqiah, & Supiana, 2021). Predictions for
future events are critical information in various planning and decision-making processes.

This study aims to compare the performance of several forecasting models for
business forecasting applications. The first model is ARIMA, or autoregressive integrated
moving average, which combines autoregressive (AR) and moving average (MA)
processes and builds a composite model of the observed time series (Lince, 2022). The
second method is Prophet, a model created by Facebook. Prophet is optimised for
business forecasting observed by Facebook, such as time, daily, weekly, past data
observations, data per year, outliers, trend changes, missing observations, and non-linear
trends.

The study aimed to produce a simple dashboard that displays graphs of recorded
data and data prediction of the results of two methods: ARIMA and Prophet Facebook.
The dashboard will be designed using the R language with Shiny components to create

dashboards.
Business Intelligence (BI) can be defined as a set of techniques and tools used to

acquire and transform raw data into meaningful and valuable information for business
analysis purposes (Yahya, 2023). The difference between BI and BA (business analytics),
according to Thomas Davenport, professor of information technology and management
at Babson College, is that BI is divided into querying, reporting, online analytical
processing (OLAP), and BA focuses on statistics, prediction, and optimisation rather than
reporting. Business Intelligence and Analytics is a term that combines both concepts

(Maziyyah, 2022). Some analytical types include:
1. Decision analytics (Decision Analytics): assists in decision-making by humans with

visual analytics of user models to reflect thinking
2. Descriptive Analytics: gain insights from past data with reports, clustering, etc.
3. Predictive Analytics: using predictive models using statistical and machine learning

techniques
4. Prescriptive Analytics: recommend decisions with optimisation, simulation, etc.

Research Methods
Data Preprocessing and Application Planning

The data used in the test was taken from Document Solution's company database
for 81 months every day from January 2017 to December 2022. The data retrieved has
four attributes: month, COGS, MOP, and Revenue. Each attribute is individual, and this
study focuses on univariate predictions, specifically on revenue figures.

The dashboard application will be designed using R Shiny, with R version 4.3.2. R
is a programming language commonly used in data analysis, and in this study, it was

developed into a dashboard.

Dashboard Using R Shiny For Arima and Prophet Comparison

Jurnal Indonesia Sosial Teknologi, Vol. 5, No. 01, January 2024 311

Database Planning

Figure 1 shows the star schema of the data warehouse used for dashboard design.

Figure 1 Star Schema Dashboard

Use Case Diagram

Use case diagrams are created to plan what each actor can do in the use of the
application or system to be created. Figure 2 shows the administrator use case diagram,

and Figure 3 shows the user use case.

Figure 2 Administrator Use Case Diagram

Christopher Caleb

Jurnal Indonesia Sosial Teknologi, Vol. 5, No. 01, January 2024 312

Figure 3 Administrator Use Case Diagram

User Interface

The user must obtain a username and password from the administrator, and the user
cannot change the data. Figure 4 shows the first dashboard page, which shows the KPI.
Each value can be adjusted by the date. The amount of data will also have an impact on
the test results. The sidebar shows several tabs that show the pages present in the
application. There are four pages: the KPI page, the Forecast page shown in Figure 5, the
Comparison page in Figure 6, and the data table page. Below the tab, there is a section
that accepts user input for the number of times to be predicted by each method. By default,
it is set to 30 days. Below the data graph, there are MAE and RMSE calculations, which
will change based on the distance of the first and last data and the number of days

predicted.

Figure 4 KPI Dashboard page

Dashboard Using R Shiny For Arima and Prophet Comparison

Jurnal Indonesia Sosial Teknologi, Vol. 5, No. 01, January 2024 313

Figure 5 Forecast Dashboard page

Figure 6 Comparison Dashboard page

Results and Discussion

Mean Absolute Error (MAE) and Root Mean Square Error (RMSE) are basic
metrics used for error testing and assessing the accuracy of predictive models. The MAE
measures the average absolute difference between the predicted value and the actual
value, thus directly indicating the overall error of the model (Firdaus, Putra, Arifandi,
Anam, & Lathifah, 2023). It offers a simple, easy-to-interpret measure of accuracy, where
smaller MAE values indicate better predictive performance. On the other hand, the RMSE
extends this concept by taking the square root of the mean squared difference between
the predicted value and the actual value. The RMSE penalises more significant errors,
making them sensitive to outliers and emphasising the importance of minimising
significant errors (PURWANTI, Rochim, & Warsito, 2022). MAE and RMSE serve as
valuable tools for model evaluation, allowing data scientists and analysts to measure and
compare the performance of various models, helping them make informed decisions
about the suitability of predictive models for specific tasks.

After conducting a thorough analysis and comparison of forecasting models, it can
be concluded that the Prophet's forecasting model outperformed ARIMA in terms of
accuracy, as evidenced by the lower Root Mean Square Error (RMSE) and Mean

Christopher Caleb

Jurnal Indonesia Sosial Teknologi, Vol. 5, No. 01, January 2024 314

Absolute Error (MAE) metrics. Prophet, a time series forecasting tool developed by
Facebook, has proven accurate in predicting time series data (Junus, Tarno, & Kartikasari,
2023). Flexibility in handling different types of data and the ability to capture seasonality
and holidays make it an excellent choice for time series forecasting tasks. In contrast,
ARIMA, although a popular choice in time series analysis, may struggle with complex,
nonlinear data patterns. Prophet's superior performance, as demonstrated by lower RMSE
and MAE values, confirms its efficacy as a forecasting tool in a wide range of
applications. Table 3.1 shows the RMSE and MAE for prediction data 30 days after the
last date.

Table 1
Comparison of MAE and RMSE ARIMA and Prophet

Algoritm
a

ARIMA Prophet

Data
Duration

01 January
2017-31

December 2022

01 January
2017-31

December 2022
Day

Predictio
n

Next 30 days Next 30 days

RMSE 5148516197.40
999

4208014977.615
32

MAE 3582009597.46
667

2210501626.756
46

Algoritm

a
ARIMA Prophet

Data
Duration

01 January
2020-31

December 2022

01 January
2020-31

December 2022
Day

Predictio
n

Next 100 days Next 100 days

RMSE 5134521157.16
281

2436799575.525
44

MAE 3787018641.84 1575306588.296
59

Conclusion

In conclusion, R Shiny can create simple dashboard designs. Using other available
modules or components, R Shiny can create dashboards with attractive designs and ease
of use. The results of RMSE and MAE calculations show that Prophet has the upper hand

in forecasting data. The overall conclusion is as follows:
1. R shiny can be used to create dashboards, showing calculations and graphs of results
2. ARIMA and Prophet can be used to predict time series data.

Christopher Caleb

Jurnal Indonesia Sosial Teknologi, Vol. 5, No. 01, January 2024 316

Bibliography

Aisah, Heti, Zaqiah, Qiqi Yulianti, & Supiana, A. (2021). Implementasi Kebijakan
Asesmen Kemampuan Minimum (AKM): Analisis Implementasi Kebijakan AKM.
Jurnal Pendidikan Islam Al-Affan, 1(2), 128–135.

Aksenta, Almasari, Irmawati, Irmawati, Ridwan, Achmad, Hayati, Nur, Sepriano,
Sepriano, Herlinah, Herlinah, Silalah, Ayupitha Tiara, Pipin, Sio Jurnalis,
Abdurrohim, Iim, & Boari, Yoseb. (2023). Literasi Digital: Pengetahuan &
Transformasi Terkini Teknologi Digital Era Industri 4.0 dan Sociaty 5.0. PT.
Sonpedia Publishing Indonesia.

Ella, Susy, & Andari, Rosita Novi. (2022). Manajemen Kontinuitas Bisnis Badan Usaha
Milik Desa di tengah COVID-19: Studi Kasus BUMDes Gua Bahu Desa Wisata
Kertayasa Kabupaten Pangandaran. Jurnal Kepariwisataan: Destinasi,
Hospitalitas Dan Perjalanan, 6(2), 253–272. https://doi.org/10.34013/jk.v6i2.824

Firdaus, Tengku Alvin, Putra, Rinno Hendika, Arifandi, Fajar, Anam, M. Khairul, &
Lathifah, Lathifah. (2023). Implementasi Sistem Rencana Pembelajaran Semester
Berbasis Web Untuk Mempermudah Proses Pembelajaran. Jurnal Teknoinfo, 17(1),
156–169.

Jamaludin, M., Sulistianto, S. W., MI, A., MM, M., Marthalia, Debby, Wikansari,
Rinandita, Fachrurazi, H., Hiswanti, S., Indrayani, Etin, & Khasanah, S. Pd. (2022).
Transformasi Digital Dalam Dunia Bisnis. Cendikia Mulia Mandiri.

Junus, Chea Zahrah Vaganza, Tarno, Tarno, & Kartikasari, Puspita. (2023). Klasifikasi
Menggunakan Metode Support Vector Machine Dan Random Forest Untuk Deteksi
Awal Risiko Diabetes Melitus. Jurnal Gaussian, 11(3), 386–396.

Lince, Leny. (2022). Implementasi kurikulum merdeka untuk meningkatkan motivasi
belajar pada sekolah menengah kejuruan pusat keunggulan. Prosiding Seminar
Nasional Fakultas Tarbiyah Dan Ilmu Keguruan IAIM Sinjai, 1, 38–49.

Manogari, Nathanael. (2019). Analisis Deret Waktu Menggunakan Metode State Space
Dan Aplikasinya Pada Harga Saham.

Maziyyah, Atika Afifah. (2022). Penggunaan Self-service Business Intelligence Dalam
Pengambilan Keputusan Di Umkm Pada Masa Pandemi Covid-19: Studi Kasus
Toko Sayur Keluarga Yogyakarta. Universitas Islam Indonesia.

PURWANTI, Hani, Rochim, Adian Fatchur, & Warsito, Budi. (2022). Prediksi Faktor-
Faktor Penentu Adopsi Manajemen Hubungan Pelanggan Sosial Berdasarkan Pls-
Sem Dan Jaringan Syaraf Tiruan Perambatan Balik. School of Postgraduate
Studies.

Sahrudin, Asep. (2016). Implementasi model pembelajaran means-ends analysis untuk
meningkatkan kemampuan pemecahan masalah matematika mahasiswa. JUDIKA