julialinear-regression

Linear regression using flux.jl in julia

I am looking for an alternative to MLJLinearModel for linear regression in Julia. I have read a few blogs describing flux.jl as a better and powerful package to perform ML tasks. However, I have not found a well-detailed tutorial directed specifically toward linear regression. Please do recommend either an approach to perform this task or sources which can help me understand the process in details.

Solution

  • I recommend using GLM for this job. For an example:

    using GLM, DataFrames

a, b = 2, 3;
dat = DataFrame(x=1:1000, y=a .* (1:1000) .+ b .+ randn(1000))

ols = lm(@formula(y ~ x), dat)

    which yields:

    StatsModels.TableRegressionModel{LinearModel{GLM.LmResp{Vector{Float64}}, GLM.DensePredChol{Float64, LinearAlgebra.Cholesky{Float64, Matrix{Float64}}}}, Matrix{Float64}}

y ~ 1 + x

Coefficients:
───────────────────────────────────────────────────────────────────────────
               Coef.   Std. Error         t  Pr(>|t|)  Lower 95%  Upper 95%
───────────────────────────────────────────────────────────────────────────
(Intercept)  3.0607   0.0641567       47.71    <1e-99    2.9348     3.1866
x            1.99989  0.000111039  18010.65    <1e-99    1.99967    2.00011
───────────────────────────────────────────────────────────────────────────

    You can always just get the coefficients:

    julia> coef(ols)
2-element Vector{Float64}:
 3.0607023932922464
 1.9998906641774181