Light intensity prediction (ML) using BOLT

MACHINE LEARNING

Having been around for few decades, Machine learning has evolved itself for providing predictions more precisely which play vital roles in many real time applications like industries, laboratories etc. Being a subset of artificial intelligence, machine learning is employing algorithms to study and improve data through experience. It employs statistics to make about accurate predictions.

WHAT IS POLYNOMIAL REGRESSION?

It is a special case of linear regression ( a form of predictive modelling technique to bring about relations between various variables ) where a polynomial equation is fit on the data with a curvilinear relationship between the target variable and the independent variables.

Now let us take a look at building a simple light intensity predictor using BOLT MODULE and BOLT CLOUD.

HARDWARE SETUP:

The light dependent resistor (ldr) helps to detect the presence or the level of light The resistance of a LDR depends on light intensity. At low light levels, the LDR has a high resistance. As the light intensity increases, the resistance decreases.

1) Insert one lead of the LDR into the Bolt Module's 3v3 Pin and other into the A0 pin.

2) Insert one leg of the 10k Ohm resistor into the GND pin and other into the A0 pin.

3) Power the module.

SOFTWARE SETUP:

1) Login to your bolt cloud account.

2) Your bolt device has to be linked to bolt cloud through the bolt app.

3) In bolt cloud, build your product( type: input devices, GPIO)

4) Now configure the product (give hardware pin A0 a variable name).

5) In the code block select the file extension js and type the code below.

6) In the code, the variable names can be kept to your wish for the graph viewing.

7) Finally save the product and link it to your bolt device.

OUTPUT:

• Prediction points are the ones which indicates how many more data points are for future prediction.
• The number of coefficients tells the algorithm how many elements should be present in the function.
• Frame size is the number of previous data points the algorithm uses to predict the pattern of the data.(if frame size is 3, it uses last 3 data points to predict future data points).
• More data points lead to more accurate prediction.
• The blue data points connected line represents the actually data collected from the bolt cloud which is given by the light dependent resistor over a period of time.
• To predict the future values at any time, click the predict button and two lines appear.
• The red colored line represents the prediction history and the yellow line represents the next predicted pattern.

Thus the user can be aware of the system's performance by checking the cloud data.

REAL TIME APPLICATIONS(can be used ):

• In biotechnology laboratories, maintaining the right light intensity is crucial in case of growth of cells, plants, other micro organisms etc. as well as in the sterilization of workplace using uv lights. during these cases a continuous light intensity collector and predictor helps in the long run if fluctuations occur in the light intensities according to which the technicians can predict if the optimum condition can be maintained for a duration of time or not.

• In chemical laboratories to study the pattern of various chemical compounds/elements' energy absorption and emission.( where the compound/element's light emission can be studied ).

• In weather forecasting along with temperature sensor the weather can be closely predicted.

• In industries to closely evaluate the light systems' performance. with prediction it can convey whether the systems performance is deteriorating or in its perfect operating state.