MACHINE LEARNING BASED RAINFALL PREDICTION

MACHINE LEARNING BASED RAINFALL PREDICTION

Table of Contents

TOC o "1-3" h z u HYPERLINK l "_Toc131089495" 1. Research overview PAGEREF _Toc131089495 h 3

1.1 Background of research PAGEREF _Toc131089496 h 31.2 Research problem statement PAGEREF _Toc131089497 h 31.3 Research rationale PAGEREF _Toc131089498 h 42. Research aim and objectives PAGEREF _Toc131089499 h 53. Research questions PAGEREF _Toc131089500 h 54. Research methodology PAGEREF _Toc131089501 h 64.1 Research philosophy PAGEREF _Toc131089502 h 64.2 Research approach PAGEREF _Toc131089503 h 64.3 Research design PAGEREF _Toc131089504 h 74.4 Data collection and analysis PAGEREF _Toc131089505 h 74.5 Ethical consideration PAGEREF _Toc131089506 h 85. Resources PAGEREF _Toc131089507 h 96. Critical success factors PAGEREF _Toc131089508 h 107. Project schedule PAGEREF _Toc131089509 h 10Appendices PAGEREF _Toc131089510 h 12Appendix 1: Questionnaire PAGEREF _Toc131089511 h 12Reference List PAGEREF _Toc131089512 h 14

1. Research overview1.1 Background of researchRainfall prediction is an old concept that has been used for forecasting weather in densely populated geographical regions. Weather forecasting is necessary in order to obtain quality agricultural production and achieve economical balance through sustainable GDP growth. Rainfall or weather prediction has been carried out using numerical weather methods in the past and is presently being converted to machine learning approach. AI based rainfall prediction is being widely accepted presently due to its high accuracy in rainfall detection and having advantage of storing large database in cloud servers. According to insights of Hewage et al. (2021), machine learning based rainfall prediction has advantage over numerical weather forecasting methods since it is a new and improved method and the forecasting accuracy is also high. The numerical weather prediction use statistical approach for detecting medium to dense rainfall in various geographical regions. This numerical approach has not always been accurate as per past data due to which transition towards AI based rainfall prediction system is slowly taking place in meteorological divisions. This research study proposes the development of machine learning based weather forecasting software in order to predict occurrence of rainfall in different geographical regions. The software is proposed to be developed using practically collected databases and a machine learning tool which will analyse datasets for predicting future occurrence of rainfall in different geographical regions. The transitioning towards AI based weather forecasting tool will reduce overall forecasting time and help in employing data analysts in various metrological divisions. This research proposal will elaborate on the structural framework of the AI tool that is used for predicting rainfall with high accuracy.

1.2 Research problem statementThe research issue is related to sudden weather change due to environmental abnormalities such as global warming, acid rain and poor air quality index. These drastic change in weather conditions leads to imbalance in agricultural productivity and can also led to degradation of soil texture and its output. The research issue is related to numerical weather forecasting techniques in terms of inaccurate forecasting due to erroneous numerical calculations. According to thoughts of Cho et al. (2020), artificial intelligence (AI) framework is slowly replacing conventional computation methods since the software produces accurate output at significant small time period. The statistical approach in rainfall forecasting has become obsolete in present era since the conventional mathematical methods are not sufficient enough to predict rainfall in geographical regions.

Figure 1: Market size of adoption of AI weather forecasting system in US

(Source: grandviewresearch.com, 2023)

The above illustrated figure shows overall market size of AI weather forecasting system adoption over the years. It is evident that the market size is gradually increasing for the AI based weather system in US over the years. Hence, meteorological sectors are slowly adopting the technology in replace of numerical forecasting methods for better accuracy and reducing overall prediction time.

1.3 Research rationaleThe significance of this research proposal is to analyse the artificial intelligence framework in order to understand its prediction algorithm. The significant difference between numerical forecasting methods and AI based numerical system will be analysed in this research work for concept development. According to thoughts of Schultz et al. (2021), research rationale helps in understanding the benefits of conducting research work. This proposal will shed light on AI based weather forecasting system and why it is preferred over statistical calculation methods. Weather forecasting particularly rainfall prediction is important for GDP growth of the country since agricultural output of the country is dependent on adequate rainfall limit. AI based rainfall prediction system will solve the issue related to rainfall prediction in geographical regions with the help of past dataset analysis. The stepwise analysis of the AI system from collecting dataset till predicting rainfall occurrence will be conducted in this research work in order to fulfil research objectives. This research proposal will thoroughly highlight the contribution of AI system for detecting rainfall in different regions for agricultural development purposes. The overall city structuring will be conducted using weather forecasting tools based on AI framework.

2. Research aim and objectivesAim:

The aim of research work is to develop machine learning based rainfall prediction software in order to conduct accurate weather forecasts in different geographical domains.

Objectives:

To elaborate on the concept of artificial intelligence framework related to rainfall prediction in different geographical terrains

To critically discuss major advantage of machine learning based weather forecasting techniques over numerical weather prediction models

To elaborate on major drawbacks faced during development of deep learning weather forecasting software

To provide suitable recommendations in order to improve the response rate of deep learning based weather forecasting software

3. Research questionsThe significant research questions obtained from research objectives are:

What is the concept of artificial intelligence framework related to rainfall prediction in different geographical terrains?

What are major advantages of machine learning based weather forecasting techniques over numerical weather prediction models?

What are major drawbacks faced during development of deep learning weather forecasting software?

What are suitable recommendations in order to improve the response rate of deep learning based weather forecasting software?

How can numerical weather prediction software be replaced with machine learning based rainfall detection framework?

What are the needful resources in order to develop AI based weather prediction software?

4. Research methodology4.1 Research philosophyResearch philosophy helps in providing guidelines to be followed by scholars for achieving overall completion of research work. According to thoughts of Newman, and Gough (2020), philosophy in research study guides scholars to adopt certain techniques to obtain authentic results from research analysis. This research should be conducted using positivism research philosophy since practical data will be collected and nurtured for obtaining relevant research outcomes. The positivism philosophy should be selected here so that proper dataset could be formulated which can be used by the AI software for predicting weather forecasts. Other research philosophies have not been selected here since they primarily deal with nurturing secondary data. The secondary data analysis is not required here since factual data should be used for providing solutions to research aim and objectives. Research philosophy should provide suitable steps to be followed in order to effectively complete the research work that is backed by scientific data and authentic research work.

4.2 Research approachThe research approach provides scholars with techniques to be followed in order to conduct scientific research output. The research techniques are mainly certain processes to be followed in order to fulfil the research objectives with scientific data output. According to thoughts of Maisham et al. (2019), different types of research approaches used in research work are deductive, inductive approaches. This research work should be conducted using inductive research approach since practical data will be used for solving the research questions. Scientific research approach is to be conducted here that can be achieved using inductive method of research approach. Deductive research approach has not been selected here since it deals with secondary data and hence theoretical analysis will be conducted in entire research work. Secondary data are not as authentic compared to primary data and the outcomes received from secondary research work may not satisfy all identified research objectives.

4.3 Research designResearch design provides suitable research framework to be followed in order to complete scientific research work. Data backed research can be achieved by implementing analytical research design. According to thoughts of Jafarzadeh Ghoushchi et al. (2021), different types of research design commonly used by researchers are descriptive, experimental, observational and many others. Analytical research design is selected over other design frameworks since analytical tools will be used for obtaining results from dataset related to rainfall forecasting. The analytical research design will solve the research problem areas using data analytics tools. Scientific outcomes will be obtained by using analytical research design that encourage scholars to use this research design technique. Data backed research will be obtained by using analytical research design methodology which will be authentic in nature. The scientific outcomes obtained will critically solve the research issues related to development of machine learning algorithm based weather forecasting software.

4.4 Data collection and analysisIn order to create mathematical models capable of forecasting rainfall in a specific location, machine studying rainfall predictions involve gathering and evaluating data. The procedure can be divided into the following crucial steps: The first process involves collecting information on past rainfall trends in the target area. In addition to information on other pertinent weather and environmental variables like temperature, wetness, wind speed, and topography, this can also include information on the amounts, duration, periodicity, and severity of rainfall. Weather stations, remotely sensed data equipment, and other sources can all provide this data (Jiang, and Forssn, 2022). The data must be organized for study after it has been gathered. In order to do this, the data may need to be cleaned of mistakes and outliers and converted into a structure that the algorithms for machine learning can use.

The data must then be used to pick pertinent features for the model of machine learning. Finding the variables that are most predictive and those that have the strongest correlations with rainfall can help with this. Machine learning techniques can be developed for forecasting rainfall after the data has been prepared and the features have been chosen. For this aim, a variety of machine learning methods, such as decision trees, regression trees, neural, and supported vector machines can be utilised. It is necessary to examine the models once they have been trained to see how effectively they predict rainfall. This may entail calculating variables like correctness, precision, and recall as well as comparing the projected rainfall numbers to actual rainfall readings.

4.5 Ethical considerationWhen creating and implementing machine learning-based rainfall predictions, there are a number of ethical issues to take into account: If the dataset used to train machine learning models is skewed, the models themselves may be skewed. The model might not be able to reliably simulate rainfall in other places or at other points of the year, for instance, if historically rainfall data has only been gathered from specific locations or during specific seasons of the year. Make sure the data being utilised to coach the models is varied and inclusive of the total population (Taherdoost, 2022). Data on specific persons may be collected in order to study rainfall patterns, which raises privacy issues. It's crucial to make sure that all data is stored correctly and is only utilised for what it was meant for. Understanding how machine learning models create predictions can be challenging due to their complexity and interpretability. To enable consumers to comprehend how predictions are formed and which aspects are taken into account, it is crucial to guarantee that the methods are clear. When it comes to predicting rainfall, machine learning models are susceptible to errors that might have catastrophic repercussions. It's crucial to make sure there's a properly defined procedure in place for maintaining model creators and users responsibly for any mistakes or errors that are committed. Predictions of impending rains based on machine learning may affect various populations in different ways. It's critical to make sure that the images are created with the needs in mind.

5. ResourcesFor a number of factors, having adequate tools is essential for machine learning-based rainfall prediction. The precision of the rainfall prediction algorithm can be increased with the right tools, such as high-quality data and adequate computing capacity. The administration of water resources, crops, crisis reactions, and other uses that depend on rainfall statistics can all benefit from precise forecasts. Large amounts of data can be processed by machine learning algorithms, which can then instantly provide forecasts (Wan et al. 2019). Even during times of heavy rains, scaling the model to manage growing data quantities and produce forecasts rapidly can be facilitated by having the appropriate processing resources.

For the prediction of rainfall, the various resources that are required are as follows:

Data: To acquire knowledge and produce precise forecasts, machine learning systems need a lot of data. Estimates of precipitation necessitate past weather data in addition to other pertinent environmental information, such as temperature, humidity, wind speed, and pressure.

The assets for data processing: algorithms used for machine learning demand a lot of processing capacity, particularly for big datasets. High-performance computing clusters, cloud computing tools, and specialised devices like graphic processing units can all be used in this. (GPUs).

Software tools: A variety of systems and software tools, including the Python libraries sci-kit-learn, TensorFlow, and PyTorch, are accessible for creating machine learning models. These instruments can assist with modelling, model assessment, and data preprocessing.

Knowledge: Expertise in mathematics, data analysis, and computing are necessary to create reliable machine-learning models (Khan et al. 2022). To guarantee the precision and dependability of rainfall forecasts, subject-matter knowledge in forecasting and climate change research is also crucial.

Monitoring regarding information standardization: It is crucial to guarantee that the information that is utilized to train the model is of the highest calibre, accuracy, and consistency. To eliminate any mistakes or discrepancies in the data, appropriate data gathering, cleansing, and verification processes are required.

6. Critical success factorsThe following are some examples of the crucial success elements for a project using machine learning to forecast rainfall:

Data quality and quantity: Sufficient and high-quality data are essential for building precise machine-learning models. Along with past rainfall data, this also contains pertinent weather information like temperature, humidity, wind speed, and pressure. Additionally, the information must be gathered over time with consistency and be indicative of the goal region.

Selecting the proper group of characteristics or factors that affect rainfall forecast is crucial. The characteristics ought to be pertinent to the issue beginning at grasp and should have a sizable effect on predictability (Hanoon et al. 2021).

Model selection and tuning: To increase the model's precision, choose a suitable machine learning model as well as adjust its hyperparameters.

7. Project schedule

Figure 2: Project schedule

Figure 3: Critical path

The project timeline will help to complete the research work following certain scientific steps in entire research process. The project schedule provides certain guidelines to be followed in order to achieve development of a data backed research work. According to thoughts of Freitas et al. (2020), scheduling process provides analytical research approach so that all objectives could be met in entire research work. Scheduling tools such as Project Libre are used for developing a proper research plan that will help to achieve an authentic research work. The major steps involves in the scheduling process is identifying a proper research plan followed by selecting proper research deliverables. The further steps involves selecting data set and developing an artificial neural network system for implementing data analysis. The AI based framework will help to analyse the dataset in order to obtain solutions to research objectives. The scheduling process will help to identify main steps in the entire research work. Scholars will receive proper guidelines which will help them complete the research using primary analysis techniques. Error in overall research work will be eliminated if proper project schedule is implemented in research study. The main schedules in this research proposal has been identified for clear understanding purposes.

AppendicesAppendix 1: Questionnaire1. What is your age?

a. 20-30 years

b. 30-40 years

c. 40-50 years

d. 50 years and above

2. What is your gender?

a. Male

b. Female

3. Can deep learning replace numerical weather prediction method?

a. Strongly agreeb. agreec. neutral

d. disagreee. strongly disagree4. Can AI neural network conduct accurate rainfall prediction?

a. Strongly agreeb. agreec. neutral

d. disagreee. strongly disagree5. How can we improve the AI based weather forecasting software?

a. Train the neural network with adequate dataset

b. Reduce complexity of the neural network by eliminating underlying layers

c. The dataset used for training AI needs to be authentic

d. Adequate participants should be questioned for improving the validity of dataset

Reference ListCho, D., Yoo, C., Im, J. and Cha, D.H., 2020. Comparative assessment of various machine learningbased bias correction methods for numerical weather prediction model forecasts of extreme air temperatures in urban areas.Earth and Space Science,7(4), p.e2019EA000740.

Freitas, F., Silva, F.J., Campilho, R.D.S.G., Pimentel, C. and Godina, R., 2020. Development of a suitable project management approach for projects with parallel planning and execution.Procedia Manufacturing,51, pp.1544-1550.

grandviewresearch.com, (2023), weather-forecasting-system, Available from: https://www.grandviewresearch.com/industry-analysis/weather-forecasting-system-solutions-market [Accessed on: 30-03-2023]

Hanoon, M.S., Ahmed, A.N., Zaini, N.A., Razzaq, A., Kumar, P., Sherif, M., Sefelnasr, A. and El-Shafie, A., 2021. Developing machine learning algorithms for meteorological temperature and humidity forecasting at Terengganu state in Malaysia.Scientific reports,11(1), p.18935.

Hewage, P., Trovati, M., Pereira, E. and Behera, A., 2021. Deep learning-based effective fine-grained weather forecasting model.Pattern Analysis and Applications,24(1), pp.343-366.

Jafarzadeh Ghoushchi, S., Dorosti, S., Khazaeili, M. and Mardani, A., 2021. Extended approach by using bestworst method on the basis of importancenecessity concept and its application.Applied Intelligence, pp.1-15.

Jiang, W. and Forssn, C., 2022. Bayesian probability updates using sampling/importance resampling: Applications in nuclear theory.Frontiers in Physics,10, p.1130.

Khan, K., Ahmad, W., Amin, M.N. and Ahmad, A., 2022. A Systematic Review of the Research Development on the Application of Machine Learning for Concrete.Materials,15(13), p.4512.

Maisham, M., Adnan, H., Ismail, N.A.A. and Mahat, N.A.A., 2019, November. Developing a research methodology for life cycle costing framework for application in green projects. InIOP Conference Series: Earth and Environmental Science(Vol. 385, No. 1, p. 012066). IOP Publishing.

Newman, M. and Gough, D., 2020. Systematic reviews in educational research: Methodology, perspectives and application.Systematic reviews in educational research: Methodology, perspectives and application, pp.3-22.

Schultz, M.G., Betancourt, C., Gong, B., Kleinert, F., Langguth, M., Leufen, L.H., Mozaffari, A. and Stadtler, S., 2021. Can deep learning beat numerical weather prediction?.Philosophical Transactions of the Royal Society A,379(2194), p.20200097.

Taherdoost, H., 2022. What are different research approaches? Comprehensive review of qualitative, quantitative, and mixed method research, their applications, types, and limitations.Journal of Management Science & Engineering Research,5(1), pp.53-63.

Wan, Z., Xia, X., Lo, D. and Murphy, G.C., 2019. How does machine learning change software development practices?.IEEE Transactions on Software Engineering,47(9), pp.1857-1871.